Theory & methods

Theory & methods

1. Observation & behavioral coding

   • Behavioral observation is a research method centered on the identification of behaviors worth theorizing and enables the testing of theories of behavior - live or audio
   • The video and audio records that are created in many behavioral observation studies provide one of the richest sources of information available in the social sciences for the study of social interactions
   • coding systems, the recording of observations and of codes, and statistical analyses. Each of these advances has been related to advances in computer technology
   •  Natural sampling methods (like the  EAR) will become easier and easier. Further, computerized coding of behavior without the need for human coders
   • lacking, is the accumulation of information on the reliability and validity of coding systems, beyond the focus of the field on interrater agreement
   • . Scholars may be discouraged from collecting behavioral data in part because of the complexities involved in collecting, coding But can provide insight that other methods cannot
   • What is observation
     • often observed short periods, but multiple times
     • sometimes observations are conducted “live.” More often, an audio or video recording is made
     • The final result is a sequential record of the behaviors of one or more individuals
     • different goals - exporatory, if intervention worked, interactions
     • Indicators derived from behavioral observation often are weakly correlated with self-report measures of the same constructs, and the meaning of this may be unclear
     • audio or video records - protection of confidentiality and anonymity
     • Naturalistic observation
       • Naturalistic observation has a long history in the study of animals - revelation life is mundane
       • favor this type of observation use a qualitative approach, where the coding system is not predetermined. Others use a quantitative approach, marked by the use of preset codes 
       • “Problems of rigor and systematization in observation took a back seat in the hurlyburly of simply trying to keep up with a movement that often seemed to us to be ruled by whimsy.”
       • Observing surreptitiously without modern hidden recorders also necessitated observers taking frequent bathroom breaks or walks - note taking
       • Coding
         • Each code is used to mark the occurrence of a specific behavior or set of behaviors
         • summarizes the physical and/or verbal behaviors of the participants of interest into specific categories using a clearly defined system of “codes” that are assigned based on a set of rules
         •  four categories (with codes within those categories): (1) current location (e.g., home, outside), (2) activity (e.g., watching TV, eating), (3) social interaction (e.g., alone, talking on phone, in group), and (4) behavioral indicators of mood
         • Topographical coding systems measure the occurrence of behaviors. Dimensional coding systems measure the intensity of behaviors along a dimension
         •  The choice of a coding system depends on the specific purposes of a study -  ‘Do you have a coding scheme I can borrow?’ This seems to us a little like wearing someone else’s underwear
         • It is unfortunate when researchers realize after the coding has been completed that they failed to code a critical behavior
         • This background work might lead a researcher to discover that a “just right” coding system is simply not available
         • Interrater-agreement
           • interrater agreement (i.e., independent observers applying the same codes to a given stream of behavior
           • training observers - careful training of observers  is essential to behavioral observation. People who may have very different perceptions of behavior must, through the training process, come to be interchangeable with one another
           • A manual should (a) describe the coding procedures in detail, (b) clearly spell out distinctions among behaviors, and (c) provide illustrative examples
           • The training videos should be selected to illustrate the full range of behaviors being coded, with progressively more difficult cases presented over time
           • Additionally, it is important that these “reliability videos” are selected at random, typically during each week of coding.
           • They clearly are useful and vital in training coders, where interrater agreement statistics can be used to monitor progress toward a quantitative agreement
           • Cohen's kappa
             • Cohen’s Kappa. Cohen’s (1960) kappa is probably the most widely used interrater agreement statistic
             • The scale of measurement and the variables formed from observational data largely determine the interrater agreement statistic used. 
             • categorical ratings are often summarized across an entire observation period into frequency and duration variables
             • By far the greatest limitation of kappa is how it is affected by distributional asymmetries
             • .40 to .59 are fair, .60 to .74 are good, and .75 and above are excellent
             • Weighted kappa  is an alternative to kappa that allows the researcher to penalize more heavily for some disagreements than others
             • Sequential analysis
               • Although what people do when interacting is important, how interactions unfold across time is possibly more important
               •  “the dimension of time is so central to conceptualizing social interaction that its use will lead us to think of interaction itself as temporal form.” 
               • Second, sequential analyses sometimes reveal unexpected patterns and thus serve as a theory-generating, as well as a theorytesting, tool.
               • unidirectional dependence, bidirectional dependence, dominance, structure of data
               • first,  the coding must reflect the order in which the behaviors were performed.  Second, codes must be mutually exclusive (i.e., only one code per event). Third, the codes must be exhaustive (i.e., there is a code for each behavior)
               • Loglinear methods provide a flexible alternative approach to sequential analysis
               • Transition matrix
                 • Once the transition matrix is formed, the conditional probability (i.e., the probability of a behavior being emitted, given a particular antecedent behavior) can be computed
                 • A transition matrix contains the tally of the pairs revealed by the moving window. If the events are not contiguous, a transition matrix for a specified lag can be computed
             • +
               • nonlinear difference equations – a “dynamical system” – to model change over time in couples’ behavior
               •  capture uninfluenced steady states as well as multiple types of nonlinear influence from one person to another
               • multilevel survival analysis - offer a novel approach to characterizing the course of an interaction, utilizing a variant of survival analysis for repeated events
               • The multilevel survival approach has applicability in any setting in which the time between a behavior’s occurrence whether an emotion display or some other behavior, marks a process of theoretical interest
               • Multilevel loglinear analysis further deftly handles cases with low rates of target behaviors, a common problem in observational research as estimates of sequential patterns among low-rate behaviors have greater measurement error
           • ICC
             • Intraclass correlation (ICC). The use of the ICC for interrater agreement in psychology was popularized by Shrout and Fleiss and is widely applied to ordinal, interval, and ratio scaled observations
             • (a) variance owing to differences among the subjects being observed and (b) variance owing to the observers
             • interrater agreement, hence the ICC, increases to the extent that between-subject variance is greater than between-observer variance
             • There are several different versions of the ICC, raising questions about which to use. Each is estimated in the context of ANOVA, in which variance is segmented into different parcels
             • Unfortunately, the ICC suffers a problem similar to that of Cohen’s kappa. The ICC is compromised by skewed distributions
         • Units
           • The most fundamental property of a coding system is the sampling strategy for behavior, otherwise known as the coding unit.
           • The major sampling strategies employed  in behavioral  observation are  event,  duration, interval, and time
           • molar/global (summary ratings are made for each code over a large number of potential coding units) vs molecular (systems code behavior as it unfolds over time and tend to have many fine-grained codes  that are given within each coding unit)
           • The large number of codes in many microbehavioral systems may make them inefficient to use, even with highly trained coders
           • (a)  coders can almost never get or maintain adequate interrater agreement on such a large number or codes, and  (b) the codes occur too infrequently in a limited observational period to make them all useful even if they were reliably coded
           • Microbehavioral systems tend to be topographical; global systems can be either topographical or dimensional,
         • Event coding
           •  •“Event triggered”   •Observe all the time –don’t record all the time, only when the behaviour occurs.   
           •  + not so time consuming + useful when you observe more than one person   
           •  - don't give time information unless onsets and offsets are noted - don't give information about preceding and subsequent     behaviours.   
           • Interval coding
             • •“Time triggered” Don’t observe all the time – don’t record all the time
             • a. Whole/Full/Complete interval  b. Partial interval  c. Momentary sampling   
             • + objectivity + useful when observation periods are prolonged (hours) + useful when recording behaviours which often occur   
             •    - not useful for behaviours that seldom occur - doesn’t give information about how behaviours are connected from one observation interval to another - you have to select interval small enough or you could lose information   
             • Situation/episode coding
               •  •“Situation triggered” •Don’t observe all the time – don’t record all the time    + easy to finding the target behaviour   
               • Time sampling
                 •  •“Time triggered” •Don’t observe all the time – don’t record all the time.   
                 • a.  Instantaneous sampling     Don’t observe all the time, just at a particular moment and   record what happens just then.   +useful when recording long-lasting behaviours   -doesn’t give information about behaviour sequences   
                 • 1 – 0 sampling (one – zero sampling).    + useful when several behaviours are studied (just 0 – 1) - doesn’t give information about how many times a behaviour has occurred - can’t be used when you are interested in sequences of behaviours   
           • Cross-classifying events
             • ''Event triggered" observe all the time, don't record all the time
             • + can study behaviour sequences + easy to use -  requires conceptual forethoughts - could lose valuable information not accounted for by the highly selective scheme   
       • Quasi
         • When investigators use quasi-naturalistic observations, the generalizability of behavior is of the highest concern, and investigators attempt to influence the situation as little as possible
         • rules transformed the otherwise typical environment into something close, but not identical, to the real world
         • constrained but otherwise acting naturally in their natural environment. This increases the quality of the data collected by increasing interaction but reduces generalizability slightly
       • Analogue observation
         • Although naturalistic observation might be appealing because the required inferences about generalizability are minimize
         •  (a) create environments where otherwise difficult or impossible to observe behaviors occur
         •  (b) enable observation of dynamic qualities of social interaction
         • (c) isolate determinants of behavior
         • often in naturalistic settings there are multiple causes of behaviors that are interdependent, making it difficult to isolate
         • There are a few important steps that must be taken when using confederates as social interaction partners. First, it is important to make sure that confederates are not a hidden source of variance - need to be blind
     • Unsystematic observation
       • + generate hypotheses   + you get a lot of information + minimize the risk of missing data   
       •  - time consuming   - can’t study more than one person at a time - you have to be alert   
       • Systematic observation
         •   1. Beforehand decide what behaviours to study   
         • 2. Operationalization of behaviours   
         • 3. Name your target behaviors
         • 4. create a behavior catalogue/coding scheme
         • +
           •  operationalization of behavior - Physically behaviours – socially based behaviours     - Molar behaviours – molecular behaviours     - Continuous recording – intermittent recording   - Coding onset and offset times   
           • Measures
             • •Frequencies   •Durations   
             • Latency Interresponse interval   
           • The mechanics of data recording
             •  1  Paper and pencil  +cheap  + easy to replace + easy to transport   
             •  2  Electronic recording devices   + saves time and labour + data are recorded on machine-readable form from      the start + no intermediate link is necessary + you may get additional information automatically            (frequencies, durations)   
           • Reliability and validity concerns
             •  Observer drift (an individual observer’s coding is not consistent over time)   Recommendations:   - Return periodically to the rating of training examples  - Check both intra- and inter-observer reliability  
             • Expectancy effects: observers are not blind to hypothesis
             •  Are observational findings affected by: 1. The presence of the observer 2. The type of task imposed by the observer 3. The location of the observation  
             • 1.The presence of the observer   - Level of intrusiveness     - Habituation effects   
             • Strategies in reducing reactivity •Not observing for the first 10 min  •Not observing at all at the first visit   •Ensuring that the same observer visits each time  •Using unobtrusive equipment •Familiarizing participants with recording procedures   
             •    2. The type of task imposed by the observer 3. The location of the observation   Recommendations:   Carry out observations across multiple settings and  tasks (Gardner, et al., 2000; Miller, et al., 2002).  

2. Survey & implicit measurement

   • Given the strong traditions of representative sampling in other disciplines, psychologists place themselves at a disadvantage if they completely ignore the scientific imperative to confirm the generalizability and applicability of their finding - survey data
   • Specifically, the huge and growing literature on questionnaire design offers guidelines for optimizing measurement in laboratory experiments of convenience samples
   • suboptimality is social-personality psychologists’ continued reliance on agree-disagree rating scales in the face of the huge literature documenting the damaging impact of acquiescence response bias
   • much greater in the entire adult population than in the subpopulation of students enrolled in college psychology courses
   • And as a result, our theories may end up being richer, especially thanks to the emergence of new technologies that permit collection of many sorts of data in the course of daily life from representative samples of people
   • Survey
     • Survey, refers to the interview or survey to study people's attitudes and/or opinions, and to describe and predict their behavior.
     • Survey methodology may vary and range from structured questionnaire to semi-structured interview.
     • Survey methodology is very common among social psychologists but personality psychologists have more extensive experience in  survey research as most personality research is based on surveys. 
     • Experiments are assumed to provide a more reliable answer when it comes to causality. However, with the development of methods for investigating causal relations (e.g., path-and mediation  analysis) the gap has been bridged
     • Research objectives should be well-defined before survey construction. This includes theoretical and empirical relationship between variables and predictions to be tested in the study. Well-defined research objectives ease every step in the process ahead
     • Well-defined concepts are also helpful, in fact, essential to achieve good validity  and reliability 
     • Items
       • In order to control for response set, you need to reverse nearly half of the items measuring a specific concept.
       •  “double-negative items” and “two questions in one” should be avoided. Be short and clear. Pretesting is essential here
       • open-ended, close-ended, text analysis
       • response time measurement in relation to performance testing and in order to detect faking – consider this possibility but be aware of the complexity of response time as an indicator psychological phenomenon.
       • 3
         • Data collection can be made by paper and pencil, through internet (e.g., Survey Monkey, Amazon Mechanical Turk).
         • Also, do whatever you can to achieve a high response rate. Remember that response rate depends on many factors like, misstargeted subjects, lack of rewards, repeated measurement, and lose of interest etc. 
         • Be clear about rewards and the timeframe in the general instruction of your survey and make sure to not spam participants.
         • many survey data sets are made available at no cost, which include measures of interest to social psychologists, including the General Social Survey, the American National Election Studies, and many other surveys available through archives
         • Cross-sectional
           • Cross-sectional: Attitudes to bananas among the population or the relationship between emotional instability and attitudes to bananas
         • Longitudinal
           •  Longitudinal: The evolution of attitudes to bananas in the population
         • Experimental
           •  Experimental: The effect of stress on attitudes to bananas 
           • Making possible an experimental design, you can always include primes (pictures, words, instructions, videos, or any other) as your independent variable before and after a set of questions (dependent variables). Randomize!
       • Marker item
         • When constructing items, you need to have at least one marker question for each measured concept.
         •  This marker item would be helpful when conducting factor and item analysis and would give you some indication of how other items are doing to measure your concept
         • Editing & coding
           • Editing and coding Normally, surveys generate an extensive amount of data. Be careful with editing and coding the responses.
           • Get to know your data well before doing any statistical analysis.
           • Save a copy of the original data set and make sure to do a random control of up to 10  observations. You can ask a student or a colleague to do the control.
       • Rating vs ranking
         • rating (e.g., degree of agreement) or ranking (e.g., ranking alternative responses)
         •  Rating scale format can be Likert, bipolar/ semantic deferential, forced choice, or yes-and-no.
         • Sometimes, you should consider giving participants a chance to respond “no answer” or “neutral”
         • When considering rating or forced choice, think carefully of normative and ipsative testing and how this distinction might affect your entire research objective.
         • When constructing a scale, you need to think about the order of the items. In performance testing, you might need to have some “easy” items at the beginning. 
         • Think also of possible priming effect of your initial items that impacts the responses of the following items. Here, a simple irrelevant word might stick in participants mind and make a big difference
         • Issues such as causality (and mediation) should be taken care of and scales should be ordered accordingly
         • Response time
           • avoid too long surveys - attrition and boredom - think of rewarding
           • As for response time for items, consider the possibility of having response time for scales – also here with regard to faking and performance testing
     • Sample
       • Be clear about your population and sample. Researchers employ convenience samples most of the times. Consider other options and motivate why
       • Your target sample needs to be defined early in the process as it would guide your item and scale construction and piloting.
       • Probability and Nonprobability sampling (Convenience, Snowball sampling, Judgmental sampling, Case study/ Deviant Case)
       • Pretesting
         • Piloting is essential and there are many ways of conducting it. If you are the first to measure a concept, make sure to start with a couple of small-scale pretestings and extend this to a regular pilot study before runing the main study.
         • Make sure to have both open-ended and standardized questions about your survey to ask your pilot subjects.
         • Consider doing some modifications after pretesting. Otherwise, bad items will follow you all the way to the reviewer of your manuscript . 
         • Modifications are common
         • Statistical methods
           • Beside common methods of analysis, survey data require specific statistical methods such as factor analysis, reliability analysis, path analysis/SEM, and mediation analysis.
           • Make sure to have an idea about the method you need to employ when constructing your survey.
           • Remember, many of the common statistical method are not appropriate for analyzing data based on, for example, rank ordering or forced choice.
           • Do always look for outliers.
           • Reporting - Report your procedure carefully – replications are impossible without a clear description what you did and how you did it.  Also, uncommon statistical methods need to be motivated and described well.
   • Implicit measurements
     • Two modes of thought There are two modes of thought, unconscious and conscious. These modes have different characteristics, making them differentially applicable/appropriate to use under different circumstances. 
     • Conscious thought: Thinking about something while consciously attending to it.   Unconscious thought: Thinking about something while not attending to it. Associating, reasoning, evaluating while your consciousness is thinking about something else.
     • associative network - The structure of our memory is of course central in this context
     • Research shows that activation of information/ stereotypes/ attitudes/ from our memory occurs automatically, quickly and without conscious control. - By priming - Presence of social objects - Wearing a heavy life jacket - Walking slowly
     • Background
       • the validity of self-report measures is often challenged when people are unwilling or unable to provide accurate reports of their own psychological attributes
       • However, the evidence that has been gathered so far suggests a more complex relation between the two types of measures
       • social desirability and introspective limits may play a role for dissociations between explicit and implicit measures,
       • researchers should be careful to avoid the fallacy of reverse inference by interpreting any dissociation in these terms
       • theoretical interpretations of measurement dissociations should be supported with relevant empirical data. Such data will provide deeper insights into why implicit and explicit measures show different antecedents and correlates; advance the development of new
       •  In socially sensitive domains, for example, responses on self-report measures are often distorted -  unintentional, resource-independent, unconscious, or uncontrollable.
       • Priming methods
         • - priming (Without awareness (pictures/words) Without attention (memory task), - impression formation (Read a story about a person), -impression expression (Judge that person) 
         • implicit association test
         • Sequential priming
           • researchers adopted sequential priming tasks from cognitive psychology to study the automatic activation of attitudes
         • IAT
           • implicit association test (IAT), which stimulated the current surge in the use of implicit measures, the toolbox of available measurement instruments has grown substantially through the development of new paradigms
           • One of the most frequently used paradigms: The IAT consists of two binary categorization tasks that are combined in a manner that is either compatible or incompatible with the to-be-measured psychological attributes
           • The basic idea underlying the IAT is that quick and accurate responses are facilitated when the key mapping in the task is compatible with a participant’s preference
           • A typical IAT includes a total of five blocks. Two of the five blocks contribute the critical trials for the calculation of the so-called IAT score
           • IAT is a very flexible task that can be used to assess almost any type of association between pairs of concepts
           • reliability estimates like self-report - common concern is that the task structure of the IAT is inherently comparative, internal consistency
       • Prediction
         • same behavior showed stronger relations to implicit measures than explicit measures when processing resources were depleted
         • behavior showed stronger relations to explicit measures compared with implicit measures under conditions of unconstrained processing resources
         • explicit measures are better predictors of behavior for people with a preference for rational thinking styles, whereas implicit measures are better predictors of behavior for people with a preference for intuitive thinking styles
         • indirect measures are assumed to tap into earlier, acquired implicit representations, whereas direct self-report measures are claimed to capture more recently acquired, explicit representations
         • In addition to content-related confounds, dissociations between implicit and explicit measures may also be the result of structural task differences
         • implicit measures are highly sensitive to contextual influences, which challenges the idea that implicit measures provide context-independent assessments 

3. Brain imaging

   • We have provided a survey of the neuroimaging methods used in social and affective neuroscience, and a discussion of eight of the current controversies
   • Another aim was to provide thorough explanations of the major ongoing controversies and questions in the field that would be satisfying for those outside it and viewed as even-minded by those inside
   • In a word, we see the theme of the future of the field as expansion: in theoretical scope, in methodological and statistical depth, in breadth of impact across fields, and in size
   • , brain-as-predictor design and Bayesian meta-analysis - social neuroscience
   • developing more powerful statistical and computational methods such as multilevel modeling and structural equation modeling for neuroimaging data
   • relationships, function, contextual modulators, connectivity, coactivation
   • Social neuroscience
     • using neuroimaging to inform social psychological theories on such topics as self-knowledge, theory of mind, mentalizing, emotion regulation, empathy, and implicit attitudes
     • gap in the level of methodological knowledge between those who practice social neuroscience and those who comprise a large part of the intended audience but are not active researchers in the field, namely socialpersonality psychologists
     • This may be in part because the rapid and recent development of the field has outpaced changes in doctoral curricula and faculty hiring in psychology departments
     • “what, how, and why” guide for consumers of social neuroscience research
     • We focus on fMRI and EEG because they are the main neuroimaging modalities for social and affective neuroscience
     • neural mechanisms, process overlap & neural systems
     • forward/reverse inference,  Spuriously High Correlations?
     • Brain scan
       • substraction design - mean average brain comparable
       • behavioral genetics and seratonin transporter
       • increased awareness of the limitations of whole-brain statistics, or at least to the limitations of the thresholding procedures necessitated by whole-brain statistics
       • For a threshold of p < .0001, the threshold r is 0.82
       • a key advantage of neuroimaging relative to other dependent measures is the ability to examine putative neural mechanisms of psychological processes
       • it can be hard to determine whether a particular voxel in one subject is located in the exact same brain region as the same voxel in another subject. And even if they are located in the same brain region, two subjects may activate the same region to differing extents for idiosyncratic reasons
       • EEG
         • measures difference on voltage between electrodes and reference electrode - 
         • ERP - sum of neurons produce electrical field that streaches outside the scalp  If multiple electrodes are investigated simultaneously, it is standard to include electrode site and each of the ERP components as factors in an ANOVA design
         • neurons must be arranged in a certain way, neurons must be synchronized in some way
         • 1. electrodes attached to the scalp correspond to different brain areas 2. to record their electrical activity
         • Among the most important aspects of EEG/ERP data collection is to collect as clean data as possible - The fewer sources of electrical activity in the testing room, the better the EEG recordings of neural activity will be
         • The selection of reference location is important because it will partially determine the shape and location of observed ERP signals. In other words, a P1 recorded in an occipital electrode using a mastoid reference is not the same as a P1 recorded from the same electrode using a nose reference
         • Before averaging, it is important to clean the data to ensure that most of the variability is attributable to psychological processing rather than extraneous sources
         • +
           •  the primary weakness of fMRI methods is their inability to track the time course of neural activity - Whereas fMRI can tell us where a process is likely to have occurred, but not when, EEG methods can tell us when a process occurred but not where
           • The electrical signals, unlike the BOLD signal, are a direct measure of neural activity at the time that activity occurred. Thus, EEG recordings contain the exact information that is lost in fMRI recordings – when the processes occur, and in what order
           • Specifically, an ERP reflects the averaged time-locked signal following neural activity - researchers often examine only one or two meaningful ERP components and average the data across multiple time points to get a single estimate of the ERP effect
           • the strength of the absolute effect, positive or negative, is meaningful. To calculate an ERP, one creates epochs (time windows that correspond to specific trials)
           • The goal of ERP data analysis is to characterize a neural response in time - The fundamental issue is that to average multiple trials into a single ERP waveform, we must assume that the processes occur at the same time point.
           • not only must it be assumed that the brain regions that produce the effect are the same for all trials and all participants, but that the timing of the brain activity must also be consistent - most, ERP studies have examined low-level perceptual process or motor responses
           • t (1) unfold in exactly the same way for all trials within subject (not even considering between subject difficulties), and  (2) that we can a priori identify the temporal onset and offset of processing
       • MEG
         • where eeg measures electrical potential meg measures magnetic fields around the scalp
         • measures the same areas as eeg, neural populations in neocortex
         • contrast to eeg - participants have to stay still, we don't care about the shape of the skull, better spatial resolution, easier to decide origins and more expensive
       • NIRS
         • consists of a number of transmitters and receivers arranged around the head
         • transmitter send light of a specific frequency that's absorbed by the blood
         • receivers then measures the amount of refleted light
         • bad spatial resolution and doesn't reach all brain areas - + tho no injections no radioactivity, need not be still, can be used on everyone
       • CAT scan
         • x-ray passed through the brain from many different directions
         • can compute 3D image of the brain
         • MRI
           • MRI scanners use strong magnetic fields, magnetic field gradients, and radio waves to generate images of the organs in the body.
           • Magnetic resonance imaging. In this respect it behaves like a small bar magnet. Under normal circumstances, these hydrogen proton “bar magnets” spin in the body with their axes randomly aligned. When the body is placed in a strong magnetic field, such as an MRI scanner, the protons' axes all line up.
           • DTI
             • Using a special MRI-sequence, water produces different signals depending on how it can move in the brain. The sequence is sensitive to water diffusion gradients.
             • Images ‘White matter fiber tracts’(nerver) - uses mri scanner
             • Water diffuses along myelinated fibers  For every voxel a tensor, with a size and direction, can be calculated. This volume of tensors is called a ‘tensor map’
             • FA value from 0-1 - 0 means that the water in taht voxel can move completely free - FA density/thickness of fiber in the area
             • PET
               • radioactive tracers can be constructed to bind to specific receptors
               • when tracers decay, we can see where the tracers were situated
           • fMRI
             • fMRI is based on the same technology as magnetic resonance imaging ( MRI) -- t uses a strong magnetic field and radio waves to create detailed images of the body. But instead of creating images of organs and tissues like MRI, fMRI looks at blood flow in the brain to detect areas of activity
             • BOLD-signal - Longer stimuli will produce correspondingly longer BOLD responses, and several stimuli presented in rapid succession will produce overlapping BOLD responses that together account roughly to one cumulative BOLD response 
             • stimuli presented too close to one another in time or in regular succession will be confounded in the BOLD response and difficult statistically to differentiate from one another. There are two main design solutions to this problem: blocked and event-related designs
             • We anticipate especially rapid accumulation of knowledge in this area in the coming years because of the high demand relative to the current information available
             • The technological design factors concern the limitations of the fMRI machine and its environment. First, fMRI scanning is expensive - more cost-effective to run fewer subjects each with more trials + within design - ecological validity? artefacts
             • Many scanning centers require participants to wear earplugs to dampen the noise, even if they are also wearing headphones
             • +
               • three considerations: statistical, technological and human factors
               • The main statistical considerations derive from the facts that all fMRI studies contain at least one within-subjects factor
               • and that the data are arranged in a time-series with a potentially strong autocorrelation (i.e., later time points cannot be assumed to be independent of earlier time points)
               • Another factor is that the variable that is measured by fMRI, the blood oxygenation level-dependent (BOLD) signal, is slowed and elongated in time relative to the stimuli used in our experiments
               • statistical power and sample size
               • Conjunction analysis is based on t-tests between conditions, as it involves examining the overlap between two or more pairwise t-tests
               • Blocked desing
                 •  Blocked designs maximize the efficiency to detect differences between conditions throughout the brain, but provide no information about the shape of the curve of the BOLD response over time, called the hemodynamic response curve 
                 • trials of the same type are grouped together into relatively long “blocks” that are usually 20 to 50 seconds each, and separated by a fixation cross “resting” baseline, various other conditions, or both
               • Rapid event-related design
                 • rapid event related design in which brief events (typically 1–5 seconds each) are presented in close succession
     • Convergences
       • Convergences in neural function between different mental processes have also been used successfully to compare theories when each offers a competing process model.
       • strength of inference based on neural convergence is limited by the specificity of brain mapping
       • For example, psychologists have debated whether we understand the emotions of others by mirroring their experience directly or by first simulating their experience in our own mind and then projecting our emotions onto them
       • This overlap between the brain systems for experiencing an emotion directly and for understanding it in others has provided support for the self-projection theory of empathy.
       • Divergences
         • Neuroscience methods can also be used to find surprising divergences in the mechanisms of various processes that would otherwise seem similar
         • emerging theme in social neuroscience is that the brain networks for processing social stimuli may be distinct from those that process nonsocial stimuli.
   • Clinical observation
     • consequences these injuries have on behaviour, we can try to the deduce the function of the damaged system/areas
     • Phineas Gage, Brocas and Wernickes area
     • electrical stimulation of the nervous system
     • neuropsychological testing: mirror-drawing task, Wisconsins card sorting test, recency memory task, verbal fluency, reading comprehension
     • Electrical stimulation
       • possible to stick electrode into the brain, increase or decrease neural activity - can study behavior from that
       • can be used clinically - stimulate cortex near damage stroke
       • Deep brain stimulation
         •  Electrodes are implanted in different places to give on-going stimulation
         • parkinson disease or epilepsy
         • research on psychiatric diagnosis - depression or schizophrenia
         • Transcranial magnetic stimulation
           • creates strong magnetic pulse, which affects neural activity
           • the magnetic field reaches through the skull and activates neurons - associations brain vs behavior
           • repetitive stimulation rtms during several minutes can give long-lasting effects - similar to deep brain stimulation
           • used to treat depression
     • Psychophysiology
       • sympathetic vs parasympatheic system - phasic vs tonic response
       • electromyography - face muscles
       • skin conductance
       • SC
         • skin conductance
         • slow - 1-2s lag after event
         • can measure both phasic and tonic responses
         • modulated by context, amygdala also active here, but the signal is affected by many more processes in the brain
         • SCR
           • a physiological response that is tied to specific tied event
           • simplest to measure and gives best quality data
           • very often used in experimental emotional learning, which is used to understand underlying processes of fear and treatment of fear
           • SCL
             • measures skin conductance over a longer period of time
             • usually to measure something that is not easily timed or tonic responses
             • for example during mental imagery of emotional story

4. Psychophysiology

   •  physiological measures provide convincing and fairly unobtrusive evidence to support testable hypotheses
   • Specific psychophysiological indexes derive their validity from psychophysiological theory confirmed via systematic empirical work.
   • Researchers must think about using physiological indexes in the same ways they think of using subjective and behavioral indexes, convincing themselves and others of the validity and reliability
   • conceptual explication of target psychological constructs, on identification of key physiological response patterns across multiple physiological measures, and on specification of appropriate situational contexts
   • Physiological indexes lend themselves to assessment of continuously fluctuating psychological states and processes, especially affective, motivational, and cognitive ones, unlikely to usefully specific attitudes, beliefs, and dispositions
   • Because physiological indexes are objective and covert, avoidance of self-report bias presents one of the major advantages of physiological indexes -  assess psychological states & processes  with other types of measures, self-report and behavioral ones or while research participants engage in other activities
   • Background
     • The value of these measures largely derives from avoidance of problems stemming from subjective self-report and human observational measures, such as participant and observer biases.
     • Social psychologists often successfully test hypotheses using validated neurophysiological response patterns associated with important psychological constructs and processes instead of, or in addition to, more traditional self-report and behavioral index
     • The methodological tools to distinguish physiologically among superordinate psychological states enable researchers to critically evaluate models
     • the promise of physiological indexes drove remedies for problems inherent in the integration of social psychology and psychophysiology.
     • Social psychophysiology
       • The advantages of social psychophysiological method accrue from its stipulation of relatively unbiased, real-time indexes of psychological processe
       • related to motivational, affective, and cognitive constructs – ones that are often difficult to quantify without bias or artifact in the absence of such objective assessments and impossible to quantify, especially temporally
       • Invariance defines the ideal relationship between a construct and its index - nominal, ordinal and interval-ratio level
       • Physiological invariants of psychological constructs have proven difficult to establishs. First, invariant indexes, whether subjective or objective, of social psychological constructs often prove elusive because the target constructs  one-to-one correspondence between specific psychological constructs and unitary physiological responses rarely exists
       • 4
         • 1. The psychological constructs, Aψ and Bψ are conceptually distinct  (One cannot index a psychological construct until one defines it explicitly and operationalizes it validly)
         • 2. The sets of physiological responses, Aϕ and Bϕ, are each more inclusive (defining a physiological index, given the many interrelationships among psychological and physiological processes, sets including two or more) more responses are considerably more desirable
         • 3. The sets of physiological responses, Aϕ and Bϕ, overlap in substance but not in form (Stronger inference results when differing internal patterns of a common set of physiological responses, rather than different sets of physiological responses, distinguish the psychological constructs)
         • 4. The sets of physiological responses, Aϕ and Bϕ, are assessed continuously over time (Continuous time-series assessment of physiological responses increases the likelihood of distinguishing differential patterns of responses between sets of overlapping responses above and beyond that of single time-point samples)
       • Basic physiological processes
         • The body can be described from a cybernetic systems or subsystems perspective dichotomizing bodily systems, albeit fuzzily, into control and operational ones
         •  the classic structures and substructures of the nervous system, such as the central nervous system, autonomic nervous system, and somatic nervous system, do not operate autonomously
         • cellular processes - Single-cell neurons provide the basic building blocks of the nervous system
         • The endocrine system functions directly to regulate growth and maturation and indirectly to modulate neural control of various operational systems.
         • Validational research
           • To validate these indexes, we conducted three types of studies: correlational, experimental, and manipulated physiology
           • correlational: Threatened participants exhibited the less benign pattern, including increases in cardiac activity and increases in total systemic vascular resistance.
           • experimental: Physiological responses were recorded continuously during the task and during a quiet rest period preceding the task.
           • manipulated physiology: Two different patterns of physiological activation, each having a distinct physical mode of elicitation, were employed. 
           • the results of the studies described in the preceding section do not exclude the possibility that physiological responses drive the psychological responses
           • Consequently, we conducted predictive validational studies to determine if the physiological indexes were predictive of future performance
           • Nonpsychological manipulations of the different patterns of cardiovascular responses themselves did not produce differences in psychological motivational states, affirming the theoretical rationale underlying the indexes
           • Constructs & context
             • in the past, explicit definitions of positive and negative affect proved no less daunting than definitions of motivational constructs such as challenge and threat, although arguably more theoreticians agree on definitions of positive and negative affect than on the definitions of moods
             • Most indexes, nonphysiological as well as physiological, of affect rest on the idea of commonalities among the specific positive emotions as a group
             •  In the development of physiological indexes of positive and negative affect, researchers have generally limited the empirical context to situations requiring relatively little movement and involving passive receipt of stimuli or information by individuals
             • These empirical limits, however, do not preclude the extension of physiological indexes of affect to more psychologically and physically active situations,
             • One-to-one relationship
               • The failure to distinguish between different patterns of cardiovascular responses associated with positive and negative motivational states in the past propelled both a large literature
               • challenged the validity of physiological indexes of affect valence such as skin conductance or heart rate changes because of a lack of one-to-one relationship with affect constructs
               • cardiovascular changes, increases in heart rate and blood pressure  used by different researchers to index oppositional motivational states, creating a dilemma for those interested in using cardiovascular indexes to identify distinctively one or the other superordinate motivational states
       • Specific methodological concerns
         • threats to validity: maturation, testing (habituation, iatrogenic changes inthe nature of the target variable), instrumentation (susceptibility to noise)
         • nr.2 solution involves either decreasing the frequency of such intrusive measurement procedures, allowing underlying tissues to recover fully between measurements, or the use of minimally intrusive monitoring equipment
         • design implications - The multiple or continuous within-participant measurement strategy proves advantageous for the indexing strategy already delineated and allows the investigator to statistically minimize individual differences in basal physiological response
         • PHYSIOLOGICAL INDEXES OF PSYCHOLOGICAL CONSTRUCTS
           • if valid and reliable psychophysiological indexes of critical constructs for testing theory in social psychology did not exist.
           • Recall that fuzzy or implicit definitions of many psychological constructs impede the successful development of any indexes of them, including psychophysiological indexes
           • challenge results from the evaluation of resources as meeting or exceeding demands, and threat results from the evaluation of demands exceeding resources
           • context: Recall also that one can develop strong psychophysiological indexes of a construct by narrowing or limiting the psychosocial context
           • Of course, limiting the psychosocial context of psychophysiological indexes precludes the possibility of invariant indexes.
           • Provided the context represents a large or meaningful cross-section of the behavioral domain, one can aim for a marker, thereby providing the basis for strong inference within the specified type of context
           • EMG
             • (EMG) recordings specific to the facial muscles associated with the expression of affect, and on the inhibition-facilitation of reflexes (i.e., startle eyeblink responses) during the experience of affective states
             • value of using electromyograms specific to targeted facial muscles as physiological indexes of positive and negative affect.
             • “somatic nervous system is the final pathway via which people interact with and modify their physical and social environments”
             • corugator supercilii EMG increases and zygomaticus majori EMG decreases during negative affect, corrugator supercilii EMG decreases and zygomaticus majori EMG increases during positive affect
             • “facial EMG can mark the valence and intensity of transient and specific affective reactions even in the absence of emotional expressions that are noticeable, at least under normal viewing conditions”
             • Startle effect
               • hypothesized: reflexes associated with positive affect  enhanced during a positive emotional state and that reflexes associated with negative affect enhanced during a negative emotional state
               • The startle eyeblink reflex is negatively toned and, hence, should be enhanced during ongoing negative affect and inhibited during ongoing positive
               • Lang and his colleagues focused on the startle eyeblink reflex, the reflexive blinks that occur when individuals perceive an unexpected and relatively intense stimulus, particularly –, but not limited to – an acoustic stimulus - emg technology required
               • Although the startle eyeblink reflex measure of Lang and his colleagues has enjoyed increasing popularity among psychophysiologists, this index has attracted the attention of social psychologists more slowly
               • Nevertheless, we urge researchers to consider it, as the startle eyeblink reflex index is relatively simple to employ and its validity is based on a wealth of research.
     • Technological background
       • Application of the sensors connecting participants’ bodies and electronic physiological recording equipment requires technological expertise, such sophistication lies within the grasp of social and personality psychologists, and its value far outweighs its cost. 
       • focus on an overview of “dry” technology,  electrophysiological recording, does not devalue the worth of “wet” technology, biochemical analysis within the endocrine-based social psychophysiological approach
       • Historically, psychophysiologists have focused their technological expertise primarily on noninvasive measurement techniques, which are ways of recording internal physiological responses of interest from the surface of the body
       • The success of this approach stems in no small part from the fact that various physiological events or responses produce detectable signals on the skin surface of the body, including both electrical and nonelectrical ones
       • The former include actual changes in electrical potentials as a function of specific physiological processes (changes in electrical potential across the heart as it completes its cycle, changes in muscle action potentials as muscle bundles contract).  The latter include changes in pressure (e.g., intra-arterial blood pressure, intraocular pressure), movement (e.g., heart, lungs, digestive organs, blood flow), temperature, and tissue
       • The path of physiological response signals for measurement purposes leads from the surface of the body to recording - transduction
       • Electrodes are used as sensors for electrical physiological signals - Although varied in size, shape - Sensors for nonelectrical physiological signals include electrodes and other devices.
       • Transducers
         • As mentioned earlier, nonelectrical physiological response signals (e.g., blood flow, sweat levels) must be converted or “transduced” to analog electrical (i.e., voltage) signal
         • This occurs early in the signal path. Most transducers operate according to principles of electrical bridging.
         •  Bridge circuits produce a continuous voltage signal representing bioelectrical physiological responses measured with electrode sensors such as skin conductance and thoracic impedance
         • a bridge circuit will similarly produce a voltage analog signal corresponding to the underlying physiological cause of the nonelectrical physiological responses (e.g., movement, temperature, blood flow)
         • sensors - transducers - conditioning - recording
     • The power of psychophysics
       • Quantitative measurement of sensory systems/ neurophysiology: – Sensory limits – Modelling, Inferring neuronal mechanisms – Diagnostic tool – Assessment tool
       •  Quantitative measurement of other mental states: – E.g. Stress, mental load, emotional states, aesthetic impression etc  
       • Provide a normative model – Signal detection theory
       • Stevens power law (1957)  Stevens took issue with Fechner's Log law. The log function is only one of many compressive functions, and Stevens argued that the correct mathematical relationship between stimulus intensity and sensory magnitude was described by a Power Law: 
       • Stevens power law P=k·Sa log(P)=K+a·log(S) Linear relations are obtained by taking the logarithm of Stevens law, with a = the slope! 
       • why is there no ideal threshold? Threshold varies from trial to trial, Random neural firing (Thurstonian noise), Stim-magnitude varies between trials, Neural quantum theory (Stevens)  
       • Brief history
         • Ernst Heinrich Weber (1795–1878) - Was one of the first to study human response to a physical stimulus in a quantitative fashion
         • Fechner, along with Wilhelm Wundt and Hermann Helmholtz, is one of the founders of modern, experimental psychology • Quantified the sensations evoked by physical stimuli: a connection between matter and mind
         • Fechner’s methods of scaling, the Weber law and the Fechner law are still widely used! 
         • Weber's law
           •  The smallest change in a stimulus that can be perceived, the just noticable difference (JND), is proportional to the initial stimulus – And also the perceived change in a stimulus is proportional to the initial stimuli
           •  Is the oldest and most well-founded "law" in psychology! • The idea that, to obtain 75 % correct discriminations, stimuli must differ by a constant percentage; not a constant amount
           •  is approximate and invalid at extremes of a range of intensities
           • Webers law still keeps researchers busy. E.g. paper in Nature Neuroscience 2019, “Now, scientists have discovered a new psychophysical rule that allowed them to identify a unique and robust explanation of Weber’s law” citation from:
           • Psychophysical methods
             •  Classical psychophysics - Absolute thresholds - detection thresholds, ”is anything there?”   - difference thresholds, ”is there any difference?”
             • -Scaling  Estimation of stimulus intensity. ”How much?”
             • Multidimensional scaling The Stockholm school “Along which dimensions do you judge the similarity of two stimuli?” i.e. color space
             •  Signal detection theory: - Measuring sensitivity (d’) AND criterion (C) independently - Denies the existence of absolute thresholds (and denies the existence of subliminal perception!
             • Near-threshold stimuli (Weak stimuli) • Thresholds (is anything there? is there any difference?)
             • Supra-threshold stimuli (Strong stimuli) • Scaling (how much?) • The relation between stimulus strength and experienced magnitude
             • Scaling methods
               • • Magnitude estimation – estimate the sensory magnitude with a number
               • • Ratio estimation – estimate with a number how much stronger a test stimulus is compared to a standard
               • • Ratio production (the inverse of ratio estimation) – eg. produce a stimulus that is for example twice as strong
               • • Magnitude matching (adjustment) – With a number: adjust a stimulus to match a number (of intensity, weight, etc) – With standard and test: adjust a stimulus (or pick a stimulus) to match another stimulus to reach a PSE (point of subjective equality) 
               • Absolute thresholds
                 • YES-NO paradigm: • Respond ”Yes” if you perceive anything or perceive a difference, ”No” otherwise • Results in bias (how liberal or conservative a person is answering Yes) • Proportion false alarms (FA) at catch trials gives an estimation of guessing frequency => an estimation of bias 
                 • Two-alternative-forced-choice (2AFC) paradigm: • 2 stim is presented, one has to be chosen, guessing if necessary • Liberated from the above mentioned bias • Other biases, or systematic errors exist: time-order effects, ex. recent is more, and left is more
                 • Fechner’s three methods: • method of constant stimuli (the only method that finds an estimate of the shape of the psychometric function, i.e. threshold AND slope) • method of limits • method of adjustment
                 • Later methods: • staircase method • adaptive testing
               • Constant stimuli
                 • 1. Select a range of stimuli from very hard to relatively easy 
                 • 2. Pick a few stimuli uniformly in this range. This will be the constant stimulus set (might include catch trials). Decide which response format to use. For example: ”Yes/No”
                 • 3. Test each stimulus many times (20-25) in random order (typically ≈ 100 trials for each threshold estimation) 
                 • 4. Present the stimuli one at a time (≈300ms)
                 • 5. Calculate the % of “yes” and “no” responses at each light level, or % correct choices
                 • 6. Fit the psychometric function
                 • Bias
                   • Number of Yes responses decreases as the proportion catch trials increases, i.e. avoiding guesses as p(Signal) decreases
                   • Higher criterion for responding Yes => Bias
                   • This offsets the threshold toward higher levels
                   • • The relation between sensitivity (Hit rate) and och specificity (False alarm rate) is important
                   • ROC
                     • • Called the ROC curve (receiver operating characteristic)
                     • • This relation can be deduced from the assumption of a threshold, and that P(obs hit)=P(true hit) + P(happened to guess right)
                     • • This relation also provide a correction for guessing formula, CFG (given that the high threshold assumption is correct) 
                     • ROC-curves under HTT are straight lines!
                     • • Plot Hits against False alarms for all criterions C (or response bias β)
                     • • Along a curve the relation between FA and Hit rates given the same sensitivity d’ but different criterion C • discriminability d’ is specified as: 1) distance between N and SN distributions d’ = Zhit – Zfa 2) or area under the ROC curve (A’), d’ = 1 correspond to A’ = 0.76
                     • HTT
                       • High threshold theory, HTT: • Assumption: All events above threshold are detected 
                       • • Threshold is influenced by trial-to-trial variability – 1- Ptrue hits is the proportion of below threshold trials when signal is present – Pobserved hits = Ptrue hits + Phappened to guess right
                       • • All catch trials are below threshold but occationally lead to false alarms due to guessing 
                       • • Pobserved hits = Ptrue hits + Pfalse alarms·(1- Ptrue hits) – In HTT: PTrue hits is independent of Pfalse alarms
                       • problems with htt: • Correction for guessing is far from perfect
                       • • Predicts linear relations between hits and false alarms (ROC-curves, Receiver Operating Characteristics), but the actual relations are often curves, not straight lines!
                     • SDT
                       • SDT: A reaction against classical psychophysics
                       • • The theory of statistical decision was translated into a theory of signal detection relevant to the detection of signals by human observers
                       • • Originally used to analyze radar operators performance during WW 2 to discriminate aircraft blobs from noise blobs on the radar screen
                       • • Decision depends on subjective criterion: e.g. how big must the blobs be to be aircraft? – Criterion differ (e.g. peace or war) 
                       • • This ability was called Reciever Operating Characteristics ROC 
                       • • Later adopted as a general psychophysical model
                       • +
                         • SDT: model assumption  • There is no absolute threshold! Rather a continuum of detectability
                         • • What is traditionally called subliminal perception occur when a high criterion is incorrectly identified as a threshold! 
                         • • There is always a noise present (with or without a signal or stimulus) 
                         • • The noise can be internal or external and varies randomly
                         • Participants are repeatedly presented the same S+N (weak Signal + Noise) and N (no signal only Noise) stimuli in random order
                         • whether x is a signal + noise (SN) or just noise (N): Sensitivity is measured by d’ (d-prime) calculated from the proportion of Hits across all SN trials, and the proportion of false alarms
                       • Criterion C
                         • • The criterion (bias) is influenced by the • number of catch trials • decision consequences (payoff) • and signal strength
                         • • Important: The criterion C (also called response bias β) does not influence the sensitivity measure d’!
                         • • C = -[Z(phit) + Z(pFA)] / 2 as measured by z-scores : C < 0 => bias toward more ”Yes” responses 
                         • C =0 => is an ideal observer criterion, no bias 
                         • C > 0 => bias toward more ”No” responses
                         • bias β is a likelihood ratio
                         • β
                           • bias β is a likelihood ratio 
                           • • β = Likelihood(SN) / Likelihood(N) = L2 /L1 = Likelihood ratio β < 1 bias toward more ”Yes” responses β = 1 is an ideal observer criterion, no bias β > 1 bias toward more ”No” responses ’
                           • • Conversion formula: ln(β) = C∙d
                           • Optimal decision criterion b: Can be calculated when: • the number of catch trials is known • payoff is known. That is, reward (e.g. money given) and punishment (money taken back) are known for Hit, Miss, False Alarms (FA), and Correct Rejection (CR) 
                           • High criterion β is used to minimize FA and maximize CR Low criterion β is used to minimize misses and maximize Hits
                           • “signal detection theory serves as a model of optimal decision making, beyond its common use as an analytic method”
                       • d-prime
                         • • useful when a researcher wishes to find out whether sensitivity is actually changed or whether a person is simply more willing to say Yes under some conditions than others
                         • • Or whenever diagnostic decisions must be made 
                         • -Do this patient have cancer? -Are there explosives in this luggage? -Will the stock market advance today? -Will an impending storm strike? -Is this person guilty? -Will this applicant succeed? 
                         • • SDT is a general science of decision-making in Yes/No situations! What started as an attempt to refine psychophysical functions turned out to have many applications beyond the exploration of human sensory ability.
                         • • Hit rate H: proportion of YES trials to which subject responded YES = P("yes" | YES) • False alarm rate F: proportion of NO trials to which subject responded YES = P("yes" | NO)
                         • • Low d’ means lots of overlap between SN and N distributions • High d’ means not much overlap between SN and N distributions 
                     • +
                       • • Threshold theories predict straight ROC lines 
                       • • ROC curves are skewed when the SN distribution is wider than the N distribution 
                       • • By some mathematical magic the percent correct in a 2AFC task is same as A’ !
         • Fechner's law
           •  He is also credited with demonstrating the non-linear relationship between psychological sensation and the physical intensity of a stimulus via the formula: {\displaystyle S=K\ln I}, which became known as the Weber–Fechner law
           • while Fechner's law is an inference from Weber's law (with additional assumptions) which states that the intensity of our sensation increases as the logarithm of an increase in energy rather than as rapidly as the increase
           • Theory: Linking scaling with thresholds The Weber & Fechner’s law
           • • Fechner realized that if every ratio in Webers law is multiplied with some constant (say 10) we should get equal differences in perceived change ΔP
           • • => if a stimulus is multiplied by a fixed factor, the corresponding perception is altered in additive constant amounts
           • This means that the relationship between stimulus and perception is logarithmic (by integrating dP = dS/S) ! • This is Fechner´s law or the Weber-Fechner´s law 0 log S S 
           • Reconciling Fechner and Stevens laws+++

5. Multilevel & longitudinal modeling

   • The techniques that we have reviewed here are extremely flexible and adaptable to allow researchers to formulate and test complex models
   • We encourage researchers, new and established, to master these techniques and to use them to enlighten the field
   •  By wedding complex models with the power of today’s multilevel and longitudinal analyses techniques, researchers can readily address the layers of interdependence and influence that characterize human behavior and its development
   • random sample from a population of interest – that is, each member of a population has an equal chance of being selected. One implication of this assumption is that observations are independent
   • In other words, the layered levels of influence can be disentangled and examined as key parameters of a statistical model
   • MLM and multilevel SEM allow researchers to examine these relationships in an accurate, parsimonious, and powerful manner
   • Nested data
     • nested data structures are no longer a statistical issue to be avoided but instead a venue to examine important theoretical relations at various levels of influence
     • With the two modern approaches to analyzing dependent data structures, researchers now have the tools to embrace the nuances of hierarchically nested data and to model the intrinsically interdependent nature of social life
     • Nested data are those in which observations are “nested” or “clustered” within groups in a hierarchical structure that is characterized, therefore, by “levels” of influence
     • Research designs that result in nested data can be broken down into two broad categories: cross-sectional designs and longitudinal designs
     • Longitudinal modeling
       • In particular, as social-personality psychologists become increasingly interested in how variables and relationships change over time (using both short- and long-time scales), longitudinal modeling
       • As is often the case, advances in methodology can stimulate richer and more nuanced research questions.
       •  Finally we turn to the analysis of longitudinal data as a special case of nested data structures (i.e., where repeated measures are nested within individuals).
       • MLM and longitudinal SEM provide the tools to not just correct for the independence violation in nested data but to systematically investigate relations between variables at two or more levels of analysis.
       • Traditionally, data from a repeated measures design were analyzed using ANOVA; however, MLM provides a more accurate and flexible framework for analyzing repeated measures designs
       •  In a repeated measures design, response (level 1) is nested within individual (level 2)
       • Panel models
         • Panel designs investigate relationships among variables over time
         • Panel designs are appropriate to use when a number of constructs have been measured at several specific points in time from a sample of individuals.
         • interested in how feelings of loneliness predict depression six months later, and in turn how levels of depression predict loneliness six months later
         •  The primary goal of panel designs is to examine predictive relations between constructs across time
         • In a panel design, all variables of interest are measured at several time points - interpreted with caution
         • When using latent variable models, however, it is necessary to ensure factorial invariance across time. Factorial invariance ensures that the latent constructs have the same definition across occasions
         • Configural invariance
           • When using latent variable models, however, it is necessary to ensure factorial invariance across time. 
           • Factorial invariance ensures that the latent constructs have the same definition across occasions; for example, that marital satisfaction is composed of the same indicators to the same degree at each occasion. 
           • weak invariance, which tests whether the relative contributions of each indicator’s variance to the construct variance
           • Three increasingly strong levels of invariance are configural invariance, which tests whether each factor is measured by the same indicators at each time;
           • strong invariance, which tests whether the relative contributions of each indicator’s mean to the construct mean are the same at each time
           • If marital satisfaction is indicated by three observed variables at the first time point, then it should be indicated by the same three observed variables
           • Weak invariance
             •  In other words, the relative contribution of each observed indicator to each latent construct is the same at each time point. T
             •  To test this, the set of factor loadings for each latent construct are set to be the same across time. Importantly, weak invariance does not assume that the latent variables have the same amount of total variance across time
             • To recap, weak invariance involves fixing factor loadings to be equal across time points, and freeing latent variances at every time point after the first
             • Strong invariance
               •  Strong invariance tests whether the indicators of a construct contribute the same relative amount to the means of the latent variables. 
               • In other words, is the pattern of higher and lower indicator means the same across time? 
               • It is not necessary to test for strong invariance when using panel models, as only variances and covariances are relevant to these models
               • Factorial invariance
       • Latent growth curves
         • latent growth curves investigate the change in one variable over time
         • One of the most popular multilevel models for longitudinal data designs is the latent growth curve mod
         •  Growth curve models, on the other hand, attempt to fit a model to the cross-time observations for each individual. 
         • These models typically include a latent intercept (i.e., a random intercept), which reflects the distribution of starting points before change has occurred, and a latent slope (i.e., a random slope of time), which reflects the distribution of rates of change over time
         • The latent intercept would reflect the distribution of marital satisfaction at the first time point (i.e., on the wedding day), and the latent slope would reflect the distribution of change in marital satisfaction over five year
         • The mean of the intercept factor is the average marital satisfaction at the beginning of marriage, and the mean of the slope factor is the average rate of change in marital satisfaction
         •  The mean of the intercept in a latent growth curve is interpreted as the mean of the dependent variable at the time point where the slope is set to 0
     • Multilevel techniques
       • Since its inception, MLM has been used to analyze data that are nested as a result of cross-sectional or longitudinal designs
       • In contrast, SEM has primarily been used to analyze data that are nested as a result of longitudinal designs (repeated measures)
       • Latent variables are constructs that are measured by several observed indicators
       • MLM can only have one outcome variable and that variable must be measured at level 1;  SEM can only accommodate one level of nesting
       • In this model, gender (at the student level) and the proportion of boys in a classrom (at the clasroom level) predicted the intercepts, with boys bullying more than girls and classrooms with higher proportions of boys having higher frequency of bullying - gender no predict
       • SEM
         • . The key differences between assessing latent growth curves in SEM and MLM are the flexibility to add complexity to the overall model that is fit to the data. In both the SEM and MLM approaches the intercept and slope of the latent growth curve can be predicted by other vari
         • . In the SEM approach, the intercept and slope of latent variables can predict and be predicted by other variables and multiple growth curves can be simultaneously estimated. 
         • In SEM, it is also possible to replace the manifest outcomes with latent variables, as seen in Figure 21.4B; this approach is sometimes called a curve-of-factors model
         • SEM can easily incorporate multiple constructs into the same model. This capability is particularly important when multiple dependent variables are of interest or when a variable acts as both a predictor and an outcome in the same model
         • SEM is a multivariate data analysis technique that can be used to investigate relationships among both observed & latent variables
       • MLM
         • MLM can be conceptualized as an extension of regression analysis that accounts for nested data and allows for effects to differ across level-2 units
         • In other words, with significant variability in the random slope of gender, the difference between men and women in job satisfaction is not the same for all workgroups.
         • When the intercept and slope vary across level-2 units, they are known as random effects; intercepts and slopes that do not vary across level-2 units are known as fixed effects.
         • this model, the level-2 predictor (department) predicts both the random intercept (β0j) and the random slope of gender (β1j)
         • In MLM, a set of con- ´ trast codes, effect codes, or dummy codes representing levels of the within-subjects factor (e.g., measurement occasions) are level-1 predictors
         • MLM allows for level-1 slopes to vary randomly across level-2 units, which cannot be accomplished using ANOVA/ANCOVA
         •  in the MLM approach, latent growth curve models are limited to a single growth curve, and the intercept and slope of a latent growth curve cannot predict other variables
       • ICC
         • An important consideration when using nested data analytic techniques is how much the nested structure of the data affects participants’ responses on the outcome variable. This cross-level influence can be quantified through the intraclass correlation (ICC)
         • The value of the ICC represents the proportion of variance in the outcome that results from differences among level-2 units
         • Small ICC values mean that a small proportion of variance is attributable to variation among level-2 units - disaggregation
         • Large ICC values mean that a large proportion of variance stems from variation among level-2 units, and if the ICC is very large (e.g. greater than .95), again a multilevel approach may not be necessary - aggregation
         • The ICC can also be interpreted as the “expected average correlation between any two randomly chosen units that are within the same group"
         • MSEM
           • advantages over traditional multilevel models, including using latent variables to represent constructs, allowing between-group variables as outcomes, modeling relations among multiple constructs at the same time, and obtaining unbiased  estimates of effects within  and  between groups 
           •  Furthermore, MSEM provides unbiased estimates of mediation effects in nested data
           • In MSEM, a structural equation model is estimated at both level 1 and level 2
           • Multilevel SEM (MSEM) is a new, flexible technique that has fewer restrictions than sem and  mlm
           • Multilevel SEM (MSEM) blends MLM and SEM to account for group-nested data, but this extension of SEM is not yet widely available.
     • Aggregation & disaggregation
       • two inappropriate methods for analyzing nested data: aggregation and disaggregation
       • each workgroup might be assigned the mean of all individual scores, and the analysis would proceed on workgroups vs individual-level data without accounting for the influence of team membership (does not take nested data into account)
       • Aggregation is the process of averaging all level-1 units within their associated level-2 - Disaggregation is the process of analyzing all level-1 units and ignoring level-2 units.
       •  Furthermore, aggregation or disaggregation can result in incorrect estimates of error variance
       • A statistical danger specific to aggregation is loss of power
       • disaggregation - making use of all observations and not accounting for their dependence, power is increased too much
     • Cross-sectional
       • When nested data are cross-sectional, researchers can use MLM to evaluate how predictors, either at level 1 or level 2, affect the outcome variable
       • how basketball players’ beliefs in collective efficacy were influenced by individual-level variables (e.g., optimism) and team-level variables (e.g., team size).
       • An important limitation of MLM is that only level-1 variables may be used as outcome variables
       • A significant indirect effect in a cross-sectional study may suggest (under extremely restrictive assumptions) a mediation effect, a spurious effect, or even a manipulation check
       • Testing mediation
         • Mediation occurs when the effect of one variable (X) on another variable (Y), is transmitted through a “mediating” variable
         • (mediation effect) is computed by multiplying a and b, and the significance of the indirect effect is assessed by testing if ab is different from 0.
         • The most popular method of testing the indirect effect is through bootstrapping, which does not assume normality of ab. When bootstrapping, the researcher assumes that his or her sample data represent a population and takes thousands of samples, with replacement, from this “population
         • By assessing mediation in a longitudinal panel model, many of the concerns about cross-sectional mediation can be alleviated.
         •  The cross-lagged paths from X to M and M to Y represent the a and b pathways, respectively, and the significance of the indirect effect is still tested with ab
         • When using the longitudinal mediation model, it is important to consider the time between measurements. There must be enough time between measurement occasions for changes in M and Y to occur
         • Either the longitudinal or half-longitudinal mediation model can be used when X is an experimental manipulation

6. Item response theory

   • estimation of latent traits (also called latent response theory)
   • IRT applied on dichotomous response categories models – correct vs. incorrect – agree vs. disagree. • Other models are possible. Multiple ordinal options
   • • Unimodality – Test should measure a single factor • Independence of items – Item responses should be independent and only associated by the examinees ability
   • – This allow us to estimate individual response patterns likelihood by multiplication of each items response probabilities – The maximum likelihood estimate provide an estimate of each individual scores
   • Traits
     • Individuals have different traits • Directly measurable: height, weight, reaction time etc
     • • Not directly measurable: IQ, math skill, verbal ability, personality, opinion, attitude etc
     • – These are hidden, or latent, but can be estimated based on the response on a series of items – This is the field of psychometry
     • CTT
       • • CTT - analyses are the easiest and most widely used form of analyses.
       • • Scoring by the number of correctly answered items or scores on a test
       • • CTT - analyses are performed on the test as a whole rather than on the item level, they apply only to that group of students on that collection of items
       • • Critique: Examinees ability is defined in terms of a particular test • Hard test => low ability (low score) • Easy test => high ability (high score) 
       • IRT+
         • • In IRT a persons underlying ability, θ, is modelled by a number typically ranging from -4 to +4 (the ability level)
         • • Item analysis provides a way of measuring the quality of questions - seeing how appropriate they were for the respondents and how well they measured their ability
         • • It also provides a way of re-using items over and over again in different tests with prior knowledge of how they are going to perform given their ability; creating a population of questions with known properties (e.g. test bank) 
         • • Test difficulty is independent of sample
         • • Examinees ability is independent of items selected in test – Different tests will give the same ability estimate for tested individuals! 
         • ICC
           • • Basic idea: Item Characteristic Curve (ICC) for a specific item in a test • The probability of obtaining a correct response as a function of persons ability (Theta θ) • Ability and difficulty is measured on the same scale! 
           • • Each item in a test inform about ability in a certain range (where each ICC have the greatestslopes), • This is shown by the item information curve (IIC) 
           • The peak of the test information curve (TIC, is the sum of all IIC´s) shows where the discriminative performance of the total test is best
           • • The ICC gives the probability that a person with a given ability level θ will answer correctly. 
           • • P(θ) = probability of correct response – As the trait level, θ, increases from left to right, the probability of correct response P(θ) increases
           • • Q(θ) = 1 – P(θ) = probability of incorrect response – As the trait level, θ, increases from left to right, the probability of incorrect response Q(θ) decreases 
           • • The test information curve should be uniformly flat and large in the range of abilities we are interested in • In certification tests, test information should be greatest around the cut-off score to accurately discriminate between proficient and non-proficient test takers
           • Vs psychophysics
             • Btw, the psychometric functions from psychophysics and ICC look similar?
             • Psychometric functions: model the relation between physical stimulus intensity and individuals probability of giving the correct answer to different items (varying item difficulty)
             • The item characteristic curve (ICC) in IRT model the relation between individuals ability and probability of giving the correct answer to a single item
             • 0. β=0, α=1 1. β=3, α=1 2. β=0, α=.4 3. β=0, α=1, c=.3  β= difficulty parameter α= discrimination parameter c= guessing parameter
           • +
             • • Is an index of usefullness of an item (measure of uncertainty of the latent trait) 
             • • Where on the ability/traitscale do the test or item discriminate best, provide most reliable estimates
             • • The one-parameterinformation curve for item j is I j (θ) = p(θ)∙[1-p(θ)] = p(θ)∙Q(θ)
             • • The two-parameter information curve for item j is I j (θ) = α? 2 ∙p(θ)∙[1-p(θ)] = α? 2 ∙p(θ)∙Q(θ) – Where α? is the discrimination parameter for item j 
             • • The item information function is at its maximum when the latent ability/trait (θi) and difficulty parameter of the item (βi) ) are equal
             • • Test information is the sum of item information curves: I(θ) = ∑I j (θ) 
             • • The standard error (SE) is a measure of uncertainty of a latent trait – The relation between standard error and test information: SE(θ)= 1/sqrt(I(θ)) – SE is dependent on the value of the latent trait θ
   • DIY
     • The first step in IRT is to create a list with examinees and responses to each item - parameter estimation
     • • Select a model to fit to the proportions of correct responses • Adjust model parameters to fit the data – for example by Maximum likelihood
     • Parameter
       • Group invariance of item parameters Two groups:
       • • Same item parameters (ex. item difficulty) will be obtained whether low or high ablity groups are used for parameter estimation! 
       • • Item parameters are properties of the item!
       • • In classical test theory item difficulty is the overall proportion of correctresponses – not group invariant – low ability group will give a higher item difficulty than the high ability group
       • Invariance
         • • Invariance of item parameters and ability is the cornerstone of IRT – Item parameters do not depend on abilities of examinees
         • – Ability of an examinee do not depend on the set of test items – For a specific examinee we obtain the same ability θ irrespectively of the item difficulty! 
         • • Test location along the ability scale doesn’t matter 
         • • An examinee can take a test that is hard or a test that is easy and obtain the same estimated ability
         • • In a classical test the examinee would get a low vs. a high test score 
         • • In IRT the examinees ability is invariant with respectto the items used in the test!
       • Estimation
         • • Ex: one correct and one wrong respons to the two items 
         • • Response curves shown in the fig A and B respectively 
         • • These curves are multiplied across θ to obtain the likelihood function (L) in fig C 
         • • L is a function of both the items and the examinees response vector  
         • • The peak of the L-curve (maximum likelihood) provides an estimate of this examinees ability θ  • The L-curve also provide an index of theerror (SEM) in estimating θ
         • When item parameters (difficulty β, discrimination α, and guessing c) are known • Maximum likelihood estimate  • Bayesian estimation procedures
         • otherwise: – The Bayesian method – Marginal maximum likelihood
         • Test construction
           • • In CTT success of item selection depends on how close the items matches the population for whom the test is intended 
           • • IRT offer a more powerful method of item selection: item parameters are invariant
           • • Item difficulty and examinees ability are measured on the same scale making it possible to select itemsthat measure over selected regions of the ability scale – For example, at a cut-off score for separating pass and fail
           • Basic approach
             • • Decide the shape of the desired test information function 
             • • Select items from the item bank with item information that fills up the area under the test information
             • • After each item is added, calculate the test information function 
             • • Continue selecting items until the area is approximately filled up (to a satisfactory degree) 
           • Applications
             • • Computer adaptive testing (CAT) 
             • – The sequence of items (avaliable from an item bank) presented to an examinee depends on the performances on earlier items in the test 
             • – Items that are maximally informative about the ability level are administred 
             • – Tests can be shortened without loss of precision (by removing items with low information at the interesting region of θ) 

7. Qualitative method

   • interested in experiences, interactions and communication in natural context
   • no hypothesis testing, generate redefine hypothesis - intepretative, diverse methods, used in several disciplines
   • quantitative vs qualitative research
   • grounded theory, IPA, discourse analysis, conversation analysis, content analysis (also quantitative) and thematic analysis (coding reliability, reflexive approaches, codebook approaches)
   • Philosophy
     • deductive vs inductive
     • objectivism (social phenomena exists independently of humans) vs  Constructivism (social phenomena/meaning is created by humans always redvised)
     • epistemology - study of nature of knowledge
     • Positivism
       • establishment of the truth
       • deduction - data tests theory -quantitative approaches
       • post-positivism: assumes reality exists, but cannot be fully understoof, findings subject to falsification (studies in post-positivism both quantitative and qualitative)
     • Constructivism
       • reality is socially constructed (through language)
       • several subjective realities
       • induction - collect data to develop theory
       • qualitative approaches
     • Pragmatism
       • observations can be converted into theories (qualitative) to then be examined through action (quantitative) and vice versa
       • "intersubjectivity" - there is a single reality but individuals can have different interpretations 
       • transferable - degree of generalization
       • goal: improve human welfare
       • adoption of research method best suited to the research question - mixed methods
       • Mixed methods
         • gather evidence based on research question and theoretical orientation
         • intentional collection of quantitative and qualitative data
         • strenght of approaches to answer research question
   • Approaches
     • data collection: informed by approach, research question, analysis approach - words, text images - diares - written text -
     • data saturation
     • inductive reasoning: interpretative, uncover or deconstruct - importance of theoretical assumptions made of overall approach - identification of thematic meanings
     • time consuming process - Software tools can help
     • Grounded theory
       • generation of theory which is grounded in the data - can discover social processes
       • simultanous data collection & analysis, categories & analytic codes from data, theoretical sampling (refine categories), discovery of social processes, memos, categories from theoretical framework
       • open coding - axial coding - selective coding
       • theory & product - explanatory framework concerning phenomena being studied
       • cyclical, iterative - researcher transparancy
     • IPA
       • interpretative phenomenological analysis
       • developed as a method to undertake experiential research in psychology
       • often complex emotional experiences, wanna understand lived experience
       • sampling is puposive, homogeneous participant group
       • in depth interview, participant leads - individually examined then considered across data
       • draws on existing theory/concepts to explore data as final step
     • Discourse analysis
       • the role of discourse in constructing social and psychological realities
       • social constructionist
       • interviews, group interviews, diares, documents
     • Types of thematic analysis
       • coding reliability: post-positivism, control for subjectivity
       • codebook: matrix or framework analysis, codebooks don't facilitate measurement, rather pragmatic, codebook used to record & chart developing analysis & guided coding
       • reflexive thematic analysis
       • Reflexive thematic analysis
         • qualitative analytic method
         • can be used in a range of theoretical & episemological approaches - theoretical informed framework, not a methodology
         • qualitative research: creative, reflexive and subjective 
         • meaning ans sense-making, context bound, interpreting, active & generative
         • assumptions: researcher subjectivity a tool, analysis & interpretations weak vs strong not accurate, quality & code themes (immersion, engagement, distancing, reflection), single or collab (not seek consensus), theme (analytic output), theme don't "emerge" from data, data analysis underpinned by theoretical assumptions
       • What is a theme?
         • something important about your data that is related to your question
         • patterned response or meaning
         • prevalence within data items/or across the data set (but not counting) - must be applied consistency
         • inductive or deductive approach to themes
         • Code vs theme
           • • Coded data differ from the units of analysis (themes), which are (often) broader
           • • Possible to code extracts of data in as many different ‘themes’ as they fit into –so an extract may be un-coded, coded once, or coded many times, as relevant
           • • A theme is “an outcome of coding... not something that is, in itself, coded” 
           • • Patterns of meaning - a shared idea or concept (central organizing concept), not summaries of meaning related to a topic 
           • • Researchers produce themes through analytic engagement with the data
           • Coding
             • • Inductive coding (and theme development) directed by the content of the data • Deductive coding (and theme development) directed by existing concepts or ideas • Semantic coding (and theme development) reflect the explicit content of the data • Latent coding (and theme development) report concepts and assumptions underpinning the data • Realist or essentialist –focuses on reporting an assumed reality evident in the data • Constructionist –focuses on how a certain reality is created by the data • Concept-driven - informed by the research literature, interview schedule topics e.g., framework analysis (Richie et al., 2003) • Data-driven - open coding
             • Descriptive Categorization Analytic
             • what's good coding: • Differing opinions!• However:• Immersion in data • Time and distance• Reflection• Interpretive depth• Codes shift and change! (however, this may be in contrast to fixed codebook approaches
             • familiarization: • Actively read and re-read the data item• Note things of interest, ideas to explore via coding• Read actively and analytically • Write notes
             • coding: • Inclusive, systematic coding –relevant to research question • Code, and re-code • Semantic or latent (latent –harder to develop) • Coding evolves
           • Theme vs topic
             • theme: Patterns of shared meaning. Central concept or ideaTelling a story
             • topic: Sometimes introduced by a question Divergent, dissimilar, diverse Summarised, not analysed
         • Semantic vs latent
         • Generating & reviewing
           • • Active and interpretive process • Patterns cross the data set • Broader level of meaning than a code
           • • Organize codes into initial themes • Use thematic maps and tables • Think about the relationship: how themes relate together • Distinctive, but part of the overall story
           • • They are reviewed and developed –return to the codes, coded data, and whole dataset • Overarching themes (organize, umbrella concept); Themes (organized around a central concept); Sub-themes – (capture an important aspect of a theme).
         • Trustwothiness
           • terms like reliability & validity are bound to quantitative research
           • credibility, transferbility, dependability, confirmability
           • discofirming case analysis

8. Motion measurement

   • • Transcranial Magnetic Stimulation (TMS)
   • • Smooth Pursuit (SP) –smooth eye movements used to track (fixate on) a moving object • Saccades –Rapid eye movements used to move gaze between objects or catch up a moving object if gaze is lagging
   • • VOR (vestibulo-ocular reflex) –smooth compensation for body movements(functional already at birth!) • Fixations –stabilizing gaze on objects
   • Mechanical variables
     • • Observation, Video recording • Motion capture • Accelerometer • Force platforms • Eye tracking (corneal reflection)
     • Electrical variables
       • • Electromyography (EMG) • Electrooculography (EOG) • Electroencephalography (EEG)
       • Metabolic/Haemodynamic variables
         • Motion capture
           • Motion capture (measures position in space)
           • • Measurement of movement can be used to study a fundamental aspect of human mental capacity: prediction of and preparation for future events
           • • Delays within the nervous system imply that we constantly base our judgments on ‘old’ information, we need to predict in order to not constantly being too late
           • •Since the method measures the actual position of the body (markers), its a good way of studying actual performance, ie the end point of several processes in the brain and body.
           • • Optoelectronic method • Active vs passive markers • What is measured: position of markers in 3D space • High spatial and temporal accuracy
           • • Issues:• The size of the measurement ’box’ • Blind spots • System calibration • The placement of the markers • Marker acceptance
         • Corneal reflection eye tracking
           • Corneal reflection eye tracking (measures gaze position in space)
           • + high temporal resolution. + Easy and fast to calibrate + Also gives pupil size (for arousal processing and more...) + Nothing attached to the subject + Easy to set up experiments +estimates gaze (great, if that is what you are interested in)
           • - Missing data issues (movement, bad lighting, bad calibration)
           • • The reflection of a fixed infrared light is captured by a high resolution/speed camera. • Algorithms calculate where the subject is looking based on the pupil and corneal reflection of the IR light.
           • Smooth pusuit
             • Object moves and the infant is stationary. In order to stabilize on the object, gaze should smoothly move with the object. This mode of control functions poorly in this 4-week-old infant
     • EOG
       • + Really high temporal resolution. + Does not cover the visual field + Allows large head movements
       • - Vertical eye movements are difficult to track - Baseline drifts - Sensitive to impedance - Needs to be complemented by head tracking (dependent on research question).
       • Eye movements are measured with EOG and head movement by motion capture.
       • In this example the experimental equipment allows independent manipulation of eye and head movements!
       • (Elektrooculogram) The eye is a dipole with a strong electrical field. This field can be measured with two or more electrodes placed near the eye (a reference and horizontal/vertical electrodes).
       • EMG
         • •Measurement reflect electrical processes within the muscles without a time delay •Intramuscular electromyography vs surface electromyography
         • •Can be used in infants, though infant motor bursts are typically “noisier” compared to adults •Modern EMG can utilize grid- or matrix-EMG to measure surfaces
         • Force platform
           • •However, forces between the body and the environment is relatively easy to measure •Force sensors are relatively accurate and inexpensive •Can be used to study (the development of) postural control
           • Kinematics vs kinetics
             • •Movement (kinematics) is created by muscle forces (kinetics), but to derive kinetics based on kinematics is very difficult (a number of parameters need to be estimated or ’guessed’) •It is also difficult to measure muscle forces directly, especially within the body.
             • •Consequently, the issue of understanding how patterns of muscle forces lead to a particular movement remains a largely unsolved problem.
             • IMUs
               • • Includes 3D accelerometers • Includes 3D gyroscope • Can include 3D magnetometer • Can be very small
               • • Allows estimation of rotation of the IMU to correct for gravity, and also improves position change estimation by combining all sensors. • Drift.

Images and files in the map