Karteikarten
MREP
Karten in diesem Stapel (126)
- 1What steps are involved in the research process ? (7)*Identify the research problem
*Literature review
*Refine the research question
*Research design and methods
*Data collection
*Data analysis and interpretation
*Report and disseminate information - 2Dissemination is?the act of spreading something, especially information, widely; circulation
- 3Quantitative research ?collects data in the form of numbers, can involve statistical analysis and hypothesis testing.
- 4Qualitative research?collects non-numerical data and seeks to understand or interpret. There is no hypothesis stated within qualitative research.
- 5Positivism*single real world reality which can be observed, measured and quantified in an objective manner
*is a philosophical theory stating that positive knowledge is based on natural phenomena and their properties and relations. Thus, information derived from sensory experience, interpreted through reason and logic, forms the exclusive source of all authoritative knowledge - 6ReductionismQuantitative research is used to answer questions relating to what, where, when, how often, how long. To answer such questions the researcher has to reduce the subject area into specific components
- 7Difference between Quantitative and Qualitative approaches Quantitative (7)Tradition of natural sciences (physics, chemistry)
Positivist paradigm
Reductionist
Objective
Deductive
Generalisable
Numbers - 8Qualitative (7)Emerged from philosophy, anthropology and other social services
Interpretivist paradigm
Holistic
Subjective
Inductive
Not generalisable
Words - 9Positivist paradigm? Interpretivist paradigm?Positivist paradigm – looks for truth or facts.
Interpretivist paradigm – interpreting or understanding. - 10Reductionist? Holistic?Reductionist - reduce aspects of a question to answer specific questions or test specific hypotheses.
Holistic – overarching understanding – looks at the whole person and not just the disease process. - 11Objective? Subjective?Objective – the processes are mechanical and involve a minimal amount of human interpretation and are therefore felt to be more objective.
Subjective – there can often be an amount of human interpretation, the research is often not detached and this is taken into account, therefore it is seen to be more subjective in nature. - 12Deductive? Inductive?Deductive – it tests the theory encapsulated in a hypothesis.
Inductive – it generates theories which encapsulates many other theories. - 13Generalisable?Generalisable - if a number of factors are considered. Qualitative is not so generalisable.
- 14Qualitative research designs (7)Grounded theory
Ethnography
Phenomenology
Life history
Case study
Discourse analysis
Narratives - 15Phenomenological method*Collect data from those who have experienced the phenomena of interest
*Analyse and “code” (identify themes) data
*Describe the “essence of the experience”
For example, I want to gain insight into the individuals experiences of having a particular health condition - 16Grounded theory method*Grounded Theory is principally concerned with why? (Process) as opposed to what or how
*Aims to generate new theory “grounded” in data.
*Discourages literature reviewing at the outset
*Sample size unpredictable
*Data collected via observation or in-depth interview
*The researcher generates mini hypothesis
*Patterns of responses begin to emerge which are then labelled/identified (Constant comparative analysis or open coding)
*Data transcribed, coded and categorized.
For example - I want to know why service users/ patients behave in a certain way - 17EthnographyEthnography is the art and science of describing a group or culture. The description may be of a small tribal group in an exotic land or a classroom in middle-class suburbia
*Approach commonly used to gather observed data
*Focuses on entire cultural groups – groups who share common behaviours and values, e.g. occupational therapists!
*Questions the impact that cultural issues have on groups,
e.g., occupational therapists behaviours
*Gather information – Field Work – May be participant or non-participant
*Identify cultural themes
*Develop a “cultural portrait”
For example, I want to gain insight into the behaviours of particular group e.g. Occupational therapists - 18Quantitative research designs (5)Randomised controlled trials
Cohort studies
Case controlled trials
Single subject experimental designs
Surveys - 19Rigour is?finding truth, reducing error
- 20The more ................... and manipulated the research, the closer we are to getting the definitive truth. However, manipulation and control may not always be possible or desirable – ethics. It is important for experiments to be rigorous as it is about reducing error and finding the truth.Controlled
- 21True experimental Quasi experimental Non experimentalTrue experimental – Randomised controlled trials – all elements
Quasi experimental – Cohort pre-post test
Non experimental – observation, study (none of 3) - 22ManipulationWhen you do something to all/part of the sample (individual or group).
‘Manipulating’ one or more variables in connection with the subjects.
E.g. Do something – give treatment or a drug to the participants. - 23ControlAbility to control or eliminate interfering & irrelevant influences from the study.
Helps to demonstrate that the end results were due to the manipulation & not through chance interference of other variables. - 24Randomised Controlled Trial (RCT) (quantitative approach)!Clinical trial – participants randomly allocated to a treatment or control group, intervention to one group compared to no intervention, follow up of both groups, gold standard for testing the efficacay of an intervention (therapy/prevention type questions) To evaluate the effectiveness of one or more forms of treatment, and is the strongest research method to find an accurate answer to interventions.
E.g. I want to know if an intervention is effective? - 25Populationall subjects taken from the general population- Inclusion/exclusion criteria – sample
- 26Randomisationgenerally all subjects allocated to one of the groups purely by chance and that every subject has an equal chance of being in either group. Groups comparable for known and unknown variables (age, sex) – any differences at outcome can be attributed to the intervention under investigation.
- 27To reduce bias what should happen?randomisation procedures should be concealed.
- 28Allocation concealment is ?whereby those involved in study do not know or have influence about who goes into which group.
- 29To reduce biasreceiving and providing should be blinded – difficult in rehab studies. Difference should only be intervention.
- 30Bias can?can deviate results away from truth or the accurate answer. RCT should minimise bias.
- 31Aspects to reduce bias/increase validity (3)*Allocation concealment
*Blinding of participants/providers of intervention
*Blinding of outcome assessor - 32Advantages of RCTRandomisation
Blinding
Considered “gold standard”
Designed to reduce bias and error - 33DisadvantagesExpensive
Limited generalizability
Volunteer bias
Ethically problematic at times
Feasibility - 34Cohort studyTo measure the occurrence of an effect and/or outcome amongst people exposed and not exposed to the suspected casual agent of interest during a specified period of follow up. The rates of occurrence in the groups are then used to support or refute the existence of an association between suspected cause and a disease or problem.
E.G.I want to know about the effects of a disease, exposure or situation - 35Advantages:Ethically safe, easier and cheaper than RCTs
Subjects can be matched
Standardisation possible
Timing, directionality and incidence can be established
Can study more than one outcome - 36DisadvantagesControls may be difficult to identify
Exposure may be linked to a hidden confounder
Blinding difficult
No randomisation
Large sample sizes or long follow-up - 37Case control study!To estimate the strength of association between an outcome or effect and a number of potentially causative factors
E.g.I want to investigate the causes of a disease (aetiology). - 38AdvantagesQuick, cheap, easy
Rare disorders or outcomes
Fewer subjects required
Variety of exposures can be evaluated - 39Disadvantages:Recruitment and selections of control groups difficult
Less well defined target population
Confounders
Potential bias: selection, recall
Only one outcome - 40Single case designsSafety and efficacy studies
Easier to complete
Allow in-depth examination of an intervention at an individual level and examination of the intervention
No control group
3 common types – AB designs, ABAB designs and multiple baseline designs - 41Aim• statement of theintent of the study
- 42Question• question regarding the specific
aspects of the topic to be investigated - 43Hypothesis• specifies what will be statistically
analysed (quantitative research only) - 44Qualitative questions are often about:Meaning
• Experience
• Practices
• Episodes
• Encounters
• Roles
• Relationships - 45Quantitative questions often ask things like:• What is the difference between X-Y?
• What is the relationship between X-Y?
• How can we best predict Z?
• How frequently does A happen?
• What causes A?
• What are the effects of A? - 46What is “an hypothesis?“A supposition or proposed explanation made on the basis of
limited evidence as a starting point for further investigation” - 47• PICO Framework for effectiveness research:Participants P
• Intervention I
• Comparison C
• Outcome O - 48Where do hypotheses come from?Inductive reasoning (i.e. observation ->
generalisation -> hypothesis)
Deductive reasoning (i.e. logically derived
from theory) - 49Null hypothesis- Null hypothesis (H0)
(the one you will be testing):
“There is no difference/ relationship...”
We can never prove an
hypothesis to be true.
We can sometimes prove an
hypothesis to be false. - 50An alternative hypothesis can be?An alternative hypothesis can be one-tailed or two-tailed:
• Two tailed (most common): used if the results can go either way
(in this case, the relationship could be positive or negative) - 51Types of hypothesesNull hypothesis (H0)
• Research/ alternative hypothesis (H1):
– Two-tailed
– One-tailed - 52A good null hypothesisMust be testable/
falsifiable! - 53How do we operationalise an hypothesis (i.e. turn a proposition into something that can be tested)?Operationalisation includes specification of:
• What is being tested (i.e. which variables)
• How are the variables tested
• Who is being tested
• When the test takes place
• The level of significance - 54P-value/alpha?the P-value must be ≤ the
critical value of 0.05 for any findings to be
statistically significant.
If P > 0.05, we would consider that any
differences are due to chance - 55Testing hypothesesOnly the null hypothesis is tested, which predicts that any
relationship/ difference is due to chance - 56The RQ (for quantitative and qualitative research) should be:• Ethical
• Relevant
• Clear and unambiguous
• Achievable
• Addressing a gap in the literature - 57For quantitative research, hypotheses should be:Testable/ falsifiable
• Clear and unambiguous, by specifying the:
– “what”
– “how”
– “when”
– “in whom”
– level of significance - 58Frameworks for Critique and Critical Appraisal ToolsCochrane Collaboration
SIGN Guidelines
Joanna Briggs
CASP: http://www.casp-uk.net/#!casp-tools-checklists/c18f8
CEBM: http://www.cebm.net/crit - 59Sources of BiasFlawed study design
Selection bias
Channeling bias
Interviewer bias
Recall bias
Citation bias - 60Accepting biasResearchers tend to accept that variables (including
bias) may impact the validity of their study
Look for sections titled under Limitations
This is typically an honest acceptance from the authors
that they may not have been able to avoid bias within
the methodology
Qualitative research commonly puts the researcher at
the centre of the process (eg. Conducting an
interview) and also in interpreting the data retrieved through this process – this is accepted.
Look for the authors statement of reflexivity when
interpreting the findings - 61Two important drivers for researchConsequence based
Rule or principle based - 62Consequence based researchMaximise the value of “benefit minus cost”
Minimising the number of participants
Minimising the cost per participant
Maximising the probability of success
Maximising the likely overall benefits - 63Fundamental ethical principlesVeracity
Truth, honesty, sincerity
Right to participants to have full information before taking part
Self-governance
Fidelity and respect
Trust
Respect right and dignity
Duty to promote well-being and autonomy
Right of self determination, privacy, respect, anonimity and confidentiality
Beneficence (Non-maleficence)
Benefit participants and society
Duty to do good and prevent harm (physical, psychological, economic)
Care and protection of the weak and vulnerable
Justice
Fairness to participants
Not giving preference to some over others (Entitlement)
Avoid discrimination, abuse, exploitation
Participants need must come before that of study - 64Unethical research practicesInvolving people without their knowledge or consent
Coercing participation
Withholding information about a study’s true purpose
Otherwise deceiving participants
Exposing participants to physical or mental stress
Invading privacy
Withholding benefits from some participants
Treating participants unfairly, without consideration, or respect
Breaching anonymity or confidentiality - 65PISParticipant Information Sheet
- 66Research involving childrenAdult considered to be aged 16+
PIS for different age ranges
5 years and under; 6-12 years; 13-15 years
AND parents / guardians
Should be shorter than for adults
Attention span; potential fear of hospitals / procedures; mental capacity if affected by disease; disease severity; previous experience of illness
Written consent from parent / guardian - 67Considering ethical issues Clear research protocolDemonstrate need for study – will be considered unethical if question has already been answered
Methodology is important – if you cannot answer the question with the methods described, the research is unethical
Get a statistician to check stats in quantitative studies
Think through the issue of consent
Think through potential risks to patients
Remember committee is about protection
PIS
Respect the process – take the time to do it properly - 68Populationan entire set of persons/ objects / events a researcher plans to study
- 69Samplea subset of the population
- 70Samplingthe selection of the sample from the population
- 71Sampling framewhere the researcher got their participants from
- 72The usual goal in sampling is to produce a representative sample(i.e., a sample that is similar to the population on all characteristics, except that it includes fewer people because it is a sample rather than the complete population).
- 73When might you sample the entire population?When your population is very small
When you have extensive resources
When you don’t expect a very high response - 743 factors that influence how representative the sample isSampling procedure
Sample size
Participation (response) - 75External validitydegree to which we can generalise the findings beyond this study
- 76External validity depends on?Depends on:
Sampling frame (where researcher got their participants)
is it appropriate for the research question & population?
Sampling strategy
probability sampling more rigorous than non-probability sampling - 77The sampling process comprises several stages:Defining the population of concern
Specifying a sampling frame, a set of items or events possible to measure
Specifying a sampling method for selecting items or events from the frame
Determining the sample size
Implementing the sampling plan
Sampling and data collecting
Reviewing the sampling process - 78Type of strategy researchers use to recruit participantsProbability sampling (random sampling)
Non-probability sampling - 79Probability sampling (random sampling)Every member of the population has an equal chance of selection
Likely to reflect the underlying population accurately - 80Non-probability samplingChance of selection for each member of the population is unequal
Less likely to accurately reflect the underlying population
Less appropriate for quantitative research. - 81Discrete dataLimited number of
options - 82Continuous dataUnlimited number of
possibilities data could
be. - 83Measurement data belong to four categories:Nominal, Ordinal, Interval and Ratio
- 841. Nominal(categorical) scaleNo relationship between data
o Gender: Male or Female.
o Crisp Flavour: Cheese & Onion, Prawn cocktail
etc
o In this type of scale you can’t say either category
is better or worse than another. Simply that they
are different. - 852. Ordinal scaleThere is an order to the data.
• For example: crisp packets are organised by size:
• Small medium Large Extra Large
• You know that small is smaller than medium, and
extra large is the biggest they have. What you don’t
know is how much bigger they are to each other. - 863. Interval scaleData are ordered with even intervals between
them. But there is no absolute zero
• For example: temperature measured in Celsius has
even interval. 10° is twice that of 5°
• However zero Celsius does not = an absence of
temperature (arbitrary point at which water
freezes). - 87Ratio scalesThis is the most sophisticated data scale. It is
ordered with even divisions between points and it
has an absolute zero.
• 15 chews per minute is 5 more than 10 chews and
20 less than 40. There is also an absolute zero = no
chewing
• Temperature measured in the Kelvin scale is
another example of a ratio scale since it has an
absolute zero = no heat. - 88Where do you get data from?Secondary Data
Primary Data
Survey/Questionnaire
Interviews - 89Secondary DataData already collected
o Published reports/audits/clinical records
o Literature search - 90Primary Data• Direct observation
o Physical measurements
• Strength, BP, Temp, flexibility, walking speed - 91Direct observation • Who makes the observation? o Self or outsider? o Issues: Bias, convenience, expense, access to data (feelings etc)Self is Cheap. No intrusion.
Subjective experience
Risk bias. ? Accuracy
? reliable
Outsider is Objective. Less bias.
Accurate recording.
Subjective experience
Intrusive
Expensive - 92Test Validity Does it capture the truth• The accuracy of the test measurement used in data
collection.
• Types of validity
o Content
o Predictive
o Sensitivity/Specificity - 93Face/Content Validity - Is it the right tool for the job• Does the measurement tool look like it can measure the
intended (face)?
• Does the tool measure all aspects (content)? - 94Predictive ValidityCan the measurement tool data predict future
outcomes? - 95Sensitivity/Specificity (accuracy)Generally applied to screening/diagnostic
tools/tests. - 96Reliability - Is the test/tool consistent? o Inter rater reliabilityBetween raters/testers
- 97Intra rater reliabilityBetween tests with
same tester - 98Data protectionGuiding principles:
o Good reason for collecting the data
o Data are accurate and contemporary
o Data are kept no longer than necessary
o Stored securely
o Not transferred to another party
o Individuals have a right to see any data which is
collected about them , this includes research
projects. - 99Framework AnalysisBecoming familiar with the data – listening to tapes, reading transcripts and studying notes.
Identifying recurring topics or issues – highlighting key words or phrases.
Organising the data into themes – developing an index of the data grouped according to themes. - 100Categorical DataCategorical data are data2
that can be classified into a number of specific categories. They are
often called nominal data. - 101Numerical DataNumerical data are data that are recorded as numerical values, but for which certain
mathematical properties hold true. There are three subgroups of numerical data – Ordinal,
Interval and Ratio data, which differ in terms of their mathematical properties. - 102Discrete dataDiscrete data are data that can only take specific or whole values. For example, the number of
children that a woman has given birth to is ratio data. It is also discrete since this can only take
whole number values – a woman can’t give birth to 2.4. - 103Continuous numerical dataContinuous numerical data can take decimal values and may be measured precisely by a machine
e.g. weight of 65.5kg. - 104Levels of MeasurementWhen talking about the kind of data we have (Nominal, Ordinal, Interval or Ratio), we usually
refer to this as the ‘level of measurement’. This is because some kinds of data are of a higher (or
better) level of measurement in terms of their mathematical properties. Nominal data have the
Introduction to Statistics for Healthcare Research Page 7 of 50
lowest level of measurement, because there is no mathematical relationship between the category
names. Next are Ordinal data, followed by Interval data and finally Ratio data are our highest
level of measurement.
____________________ - 105A pie-chart is useful fordisplaying nominal data.
- 106Bar ChartAnother graphical display for categorical data
- 107Scatter PlotsThe relationship between two numerical data variables can be assessed using a scatter plot. The
data in Figure 4 show the healing time in days (discrete ratio data) for wounds of different
dimensions (continuous ratio data). - 108outlierA data point
that is unusual in some sense when compared with the rest of the data - 109HistogramsOne of the most important graphical displays for numerical continuous data is called a
histogram. The shape of a histogram illustrates the distribution. This histogram is ‘bell shaped’ and data which display a histogram with this characteristic shape
are referred to as normal data or data which are normally distributed. - 110The main difference between a histogram and a bar chartthere are spaces
between the groups/scores along the X axis (the one along the bottom) of the bar chart but none
on the X axis of the histogram. This is because data on the X axis of a histogram are continuous
whilst in a bar chart, the data are discrete. T - 111Symmetric distribution– most of the categories are equally
represented in the histogram - 112Normal distributionshows the characteristic ‘bell’
shape with fewer data at the upper
and lower ends of the histogram
and most concentrated in the
middle (around the mean) - 113meanAverage value
- 114The median value of a set of numbers is simply thethe ‘middle’ number when the data are arranged
in ascending order of magnitude. If the number of data points is odd then this is just the middle
number. If the number of data points is even, the median is calculated as the average of the
‘middle two’ numbers. - 115modeis the value that occurs most often in a set of data.
- 116Probability sampling strategies RandomSimple
Unrestricted
Systematic
Stratified
Cluster
Multistage - 117Non-probability sampling Non-randomAccidental/Convenience
Quota
Purposive/Judgemental
Snowball
Theoretical
Volunteer - 118Simple Random SamplingEnsures that every member of target population has an equal chance of selection
Sampling frame must give us access to ALL members of the target population.
Participants selected in a truly random way
Well shuffled names out of a hat
Random selection by a computer
Random number tables - 119Systematic Random SamplingRandom start within list
Sample every ‘nth’ person from the list
(e.g., ask every 4th case on the list)
Sampling frame must give us access to ALL members of the target population. - 120Stratified SamplingSample participants from different groups in proportion to their group size
- 121Cluster SamplingRandomly sample from various ‘clusters’ which represent important sub-groups of popn. e.g area, city
Typically used when a complete list of the members of popn. is not available but a complete list of groups or 'clusters' is available
Also used when a random sample would produce a list of subjects so widely scattered that surveying them would prove to be far too expensive/impractical etc - 122Non-probability SamplingSometimes probability sampling is not required
Pilot studies
Studies where do not intend to generalise results back to the target population etc.
Sometimes random sampling is not possible
Use non-probability sampling as this is more straightforward
chance of selection for each member of the population is unequal
sample will not be representative of the wider population (but data still useful overall) - 123Convenience samplingUse the nearest, most convenient people
Least satisfactory method of sampling - 124Quota Sampling- Target number of participants for each strata in the population are set
- Sample people in relevant strata by any means necessary until obtain quota
- Sample size reflects underlying population but also element of convenience - 125Snowball samplingAsk participants to recruit others ‘like them’
For example, ask members of Age UK to recruit other older adults (though often used to get participants from fringe, e.g., gangs, drug users)
Effective way to recruit ‘hard to reach populations’
Cannot tell extent to which this sample represents the wider population
Overall, non-probability sampling is not as robust as probability sampling for quantitative research because we cannot tell if representative sample is generated - 126—
Informationen
- Beschreibung
- University MREP Karteikarten am MREP, erstellt von Morgan Morgan am 16/11/2015.
- Erstellt von
- Morgan Morgan
- Aufrufe
- 0
- Bewertung