Research methods

Page 1

Page 2

Experimental method and variables

Independent variable- what we change

Dependent variable- what we measure

Aim: the purpose for carrying out a study- 'to investigate whether' or 'to find out whether'

Hypothesis: 1. identify both conditions of the independent and dependent variable 2. explain how to operationalise the variables 3. explain how you think the independent variable will impact the dependent variable

Types of Hypotheses:

alternative- a clear precise and testable statement that states the relationship between the variables to be investigates- experimental hypothesis- directional and non-directional

null- predicts that there will be no difference or association between the variables- research is conducted to prove this wrong

directional- states the kind of difference or relationship expected between 2 conditions or groups- statistical tests require a 1 tail test as you have one outcome in mind- when there is previous research so you know what to expect

non-directional- predicts that there will be a difference or relationship between the 2 conditions- use a 2 tailed test- used where there is no previous research or previous but conflicting research

Types of experiments and variables:

lab experiments- conducted under highly controlled conditions, where accurate measurements are possible- standardised procedure- / easy to repeat, allows for precise control of extraneous and independent variables so cause and effect can be established- x low ecological validity and be open to confounding variables due to demand characteristics

field experiments- done in the everyday environment of the participants- manipulate the independent variable but cannot control the extraneous variables- / high ecological validity as participants may not know they are being studied so there is less chance of demand characteristics- x less control over extraneous variables

natural experiments- conducted in everyday environments- no control over the independent variable as it occurs naturally- / high ecological validity and less likelihood of demand characteristics plus it can be used when it would be unethical to manipulate the independent variable- x expensive and time consuming with no control over extraneous variables

quasi experiments- researcher examines the effect of a naturally occurring independent variable which is a difference between people that already exists- / high ecological validity and can be generalised- x low internal validity as there is no control over extraneous variables

confounding variables: a variable that isn't the independent variable but changes systematically with it- not controlled so affects the result

extraneous variables: a variable that doesn't vary systematically with the independent variable but may have an effect on the dependent variable- makes it hard to detect cause and effect- need to be controlled

ways to control extraneous variables:

- single blind design- participants aren't aware of the research aims

- double blind design- participants and research unaware of the aims and hypothesis

- experimental realism- researcher makes tasks sufficiently engaging so participants pay attention to the task not being observed

Page 3

Experimental design

independent measures: participants take part in 1 condition of the experiment and are randomly allocated to one of two groups- / avoids order effects- x participant variables can affect the outcome and it is time consuming- counter these issues by ensuring that participants are randomly allocated to conditions

repeated measures: participants take part in all conditions of the experiment- / saves time- x open to order effects- these can be countered using a time gap, counterbalancing or using 2 different tests

matched pairs: participants put in pairs based on a potential extraneous variable and then are put into opposing conditions- / avoids order effects, doesn't require counterbalancing and reduced participant variables- x costly and time consuming- these issues can be countered by restricting the number of variables to match participants on and by using a pilot study

Observation techniques and design

naturalistic observations: behaviour studied in natural environment and is not interfered with- / not distorted by researchers expectations, no demand characteristics, high ecological validity- x less control over variables, replication difficult

controlled observations: participants environment is controlled and there is control over some variables- easy to replicate, control of variables- x demand characteristics

covert observations: behaviour observed without participants consent- / no demand characteristics, increased validity- x less ethical, can't use results if consent isn't retrospectively given, less control of variables

overt observations: participants know they are being observed and give consent- / controlled and ethical- x demand characteristics

participant observations: observer becomes part of the study- / increased validity- x loss of objectivity

non-participant variables: researcher remains separate and objective- / objectivity- x loss of insight into the observation group

structured observations: when there's too much information target behaviours can be simplified to be easily defined- observed can quantify their observations

unstructured observations: observer writes down everything they observe- accounts rich in detail- appropriate for small groups but there may be too much happening

Sampling techniques

random sampling: a sample in which every member of the population has an equal chance of being chosen- put numbers into a hat and select the required number of names- / unbiased as all members have an equal chance of being chosen- x may end up with a biased sample

systematic sampling: using a pre-determined system to select participants from a sampling frame- / unbiased as participants are selected using an objective system- x not truly unbiased unless you select a number at random, start with that person and select every nth person

opportunity sampling: taking a sample from people who are available at the time and fit the criteria- / easy and fast- x biased

stratisfied sampling: classify the population into categories and chose a sample which consists of participants from each category in the same proportion they are in the population-/ more representative- x time consuming

volunteer sampling: participants become part of the study because they volunteer or ask to be- / access to a variety of participants meaning the sample is more representative and less biased- x biased as participants are more motivated and have more time on their hands

Page 4

Self-report techniques

analysis: questions need to be written so they can be easily analysed- closed questions are the easiest for this but they may force people to select answers which don't represent their real thoughts or behaviours

bias: should be free from bias as it could lead the respondant to give a certain answer

clarity: clear and unambiguous

- should contain filler questions to reduce demand characteristics

structured interviews: pre-determined questions- / predictable answers so easier to analyse, less interviewing skills necessary- x interviewer bias, reliability affected

unstructured interviews: new questions develop as you go- / more detailed information can be obtained- x interviewer bias, requires well trained interviewers

questionnaires:/ easily repeated, cheap and fast and people may be more willing to give personal information- x bias, time consuming

Ethical Issues

informed consent: ensures that a patient, client or participant is aware of all the potential risks and costs involved in a treatment or procedure- dealt with through: presumptive consent, providing an information sheet and consent form, and a debriefing

deception: misleading or tricking participants about the purpose or direction of the study- dealt with through a debriefing

protection from harm: must be protected from physical and mental harm- you must not embarrass, frighten, offend or harm them- dealt with through: debriefing, therapy, BPS guidelines, withdraw from study

privacy/ confidentiality: none of the participants in a study should be identifiable from any reports that are produces-dealt with through the use of initials, and storing information in a safe place

Investigator effects

- occur when a researcher unintentionally or unconsciously influences the outcome of any research they are conducting

-a way of overcoming this is through standardisation- an important part of this is standardised instructions that must be written so they can be read out and must include a check at the end to ensure that participants understand what they have to do

Page 5

Reliability and validity

validity: the extent to which a study measures what it intends to measure

internal validity: refers to whether the effects observed in a study are due to the manipulation of the independent variable and not any other factor

external validity: refers to how well you can generalise the results from the research participants to people, places and times outside the study

temporal validity: whether the results generalise to other times

mundane validity: whether the tasks reflect those of everyday life

ecological validity: how well you can generalise a study to different settings and situations

assessing validity

- face validity- whether a test, scale or measure appears to measure what it is supposed to measure 'on the face of it'- done by eyeballing the measuring instrument or passing it to an expert to check

- concurrent validity- demonstrated when the results obtained are very close to, or match, those obtained on another recognised and well-established test measuring the same concept or construct- tested by doing a correlation between the scores the participants received on both testes, if a strong and positive correlation is found it is considered to be a valid test as it has high concurrent validity

improving validity

EXPERIMENTAL RESEARCH: use a control group so the researcher is better able to assess if changes in the dependent variable where due to the independent variable- standardise procedures and instructions to minimise the impact of investigator effects on the validity of an outcome- single blind or double blind procedures to minimise the impact of investigator effects and reduce the effect of demand characteristics as participants aren't aware of the aims of the study

OBSERVATIONS: use covert observations so there is minimal intervention by the researcher leading to high ecological validity- use specific, non-ambiguous behavioural categories

QUESTIONNAIRES: include a lie scale within the questions to assess the consistency of a participant's response and to control for social desirability bias- assure participants that their responses will remain anonymous to reduce the effect of social desirability bias

QUALITATIVE METHODS: likely to have higher ecological validity than quantitative methods as they are better able to reflect a participant's reality- have to demonstrate interpretative validity of their conclusions, the extent to which the researcher's interpretation of events matches those of their participants, can be done through coherence of the researcher's reporting and direct quotes from participants

assessing reliability

reliability: a measure of consistency- refers to how consistent the findings from an investigation or measuring device are

- test re-test- used to assess the reliability of questionnaires and psychological tests and interviews- administering the same test or questionnaire to the same person on different occasions- if it is reliable the results should be the same or very similar- 2 sets of scores are correlation- > or = o.8 indicated good reliability- must be sufficient time between 2 tests but not long enough their abilities or attitudes may have changed

- inter-observer- assesses the reliability of observational research- one researcher's interpretation of events may differ which would introduce subjectivity and bias and would make the data unreliable- reconmended observations are conducted by teams of at least 2- observers discuss and agree on the behavioural categories- watch the same event(s) but record the data independently- data is then correlated

improving reliability

EXPERIMENTS: reliable as the researcher can control many aspects of the procedure more precise replication

INTERVIEWS: use the same interviewer- if this isn't possible interviewers must be properly trained so that one isn't asking questions that are too leading or ambiguous- avoided using a structured interview

QUESTIONNAIRES: if test re-test reliability is found to be low some items may need to be deselcted

OBSERVATIONS: make sure behavioural categories are properly operationalised and are measurable and self-evident- categories should not overlap and all behaviors should be covered on the checklist

Page 6

Descriptive statistics

- various ways of summarising and analysing numerical data to draw meaningful conclusions- collectively known as descriptive statistics which include measures of central tendancy and measures of spread

measures of dispersion

STANDARD DEVIATION: a single value that tells us how far scores deviate from the mean- the larger it is the more participants weren't all affected by the independent variable in the same way- the smaller it is the more data is clustered around the mean- x distorted by outliers- / more precise measure of dispersion as it includes all values

RANGE: works out the pread of scores by taking the lowest value from the highest value and adding 1 (accounts for any margin of errors)- / easy to calculate- x only considered 2 most extreme values- x unrepresentative of the data

measures of central tendency

MODE: most frequently occurring- can have no mode or be bi-modal- / easy to calculate and can be used to analyse data in categories- x not representative and can ignore important findings

MEDIAN: order numbers lowest to highest and find the middle value- / not effected by outliers and easy to calculate- x less sensitive, not all scores included in final analysis

MEAN: sum of numbers divided by amount- / more sensitive as it includes all data scores and more representative- x easily distorted and may not represent the rest of the data accurately

Levels of measurement

nominal data: data represented in the form of categories- discrete as only one item can appear in one of the categories

ordinal data: data that is ordered in some way- doesn't have equal intervals between each unit- lacks precision as it is based on subjective opinion rather than an objective measure- questionnaires, psychological tests and such don't measure something 'real' but rather measure psychological constructs- sometimes referred to as unsafe data as it lacks precision but better than nominal data

interval data: based on numerical scales that include units of equal, precisely defines size- better than ordinal data as there is more detail- anything that is measured on a public scale of measurement and so produces data based on accepted units of measurements are considered to be interval data- most precise form of data in psychology and is necessary for the use of parametric statistical tests

qualitative vs quantitative data

qualitative data: results that are expressed in the form of words and are non-numerical- / tends to have greater external validity due to providing more insight into participants- x subjective nature can lead to experimenter bias- x often difficult to analyse due to subjective nature of responses, hard to summarise statistically

quantitative data: results that can be counted, numerical data- / simple to analyse as it's in the form of statistics and more objective so it doesn't have bias in analysis- x low ecological validity due to narrow range of responses, not representative of the population

Page 7

Data distribution

data collection

primary data: information that has been obtained firsthand by the researcher for the purpose of the research project- often gathered directly as a part of the experiment, self-report or observation

secondary data: information has already been collected by someone else and so pre-dates the current research project- data located in journals, books, articles or websites

meta-analysis: refers to the process of combining results from a number of studies on a particular topic to provide an overall view- may involve a qualitative review of conclusions and/or a quantitative analysis of the results producing an effect size

data display

histogram: continuous data

scattergram: used when doing a correlation- doesn't matter which axis the variables go on

bar chart: categories on the x have no fixed order and there is no true 0- used when data is divided into categories- discrete data

tables: the measurements collected in a research study are referred to as 'raw data' which is then converted into descriptive statistics used for interpreting findings

line graph: continuous data on the x

data distribution

normal distribution: a symmetrical bell-shaped frequency distribution occurs when certain variables are measures- such events are distributed in such a way that most of the scores are clustered around the midpoint- the mean, median and mode are at the midpoint- dispertion of scores either side of the mid-point is consistent and can be expressed in standard deviations

skewed distribution: data is negatively skewed if the scored are bunched towards the right and positively skewed if they are bunched towards the left- the mode remains the highest point followed by the median then the mean- scores are not distributed evenly around the mean

correlation

correlations systematically show how strong a relationship is between 2 continuous variables

correlation illustrated the strength and direction of an association between two co-variables

positive: as one variable increases so does the other

negative: as one variable decreases so does the other

coefficients are always between -1 and 1, the bigger the number the stronger the relationship between the co-variables

near 0= zero correlation, 0.3= weak correlation, 0.5-0.79= moderate correlation, 0.8 and above= strong correlation

used in the assessment of reliability and validity- study must be >/= 0.8 to be reliable or valid