Research Methods Flashcards
Between groups design
Using separate groups of participants for each of the different conditions in the experiment.
Within subjects design
Each participant is exposed to all the conditions of the experiment.
Ratio data
Data is measured in units on a constant scale, with an absolute zero value.
Interval data
Data is measured in units on a constant scale, with equal intervals. No “zero” value.
Nominal data
When data is categorised based on an equivalent feature (i.e. name, gender, number).
Ordinal data
Ordered by rank e.g. 1st, 2nd, 3rd.
N.B. Tells us nothing about differences between values.
Level of measurement
Relationship between what is being measured and numbers obtained on a scale (i.e. accuracy of measurement.
Empirical
Gathering evidence through observation and measurement that can be replicated.
Measurement error
A discrepancy between the number we use to represent the thing we are measuring and the actual value.
What a score consists of:
1) True score
2) A score for other things we are inadvertently measuring
3) Systematic bias
4) Random error
Discrete variable
No underlying continuum. No overlapping categories. Clear beginning and end.
Continuous variable
A continuum with no clear beginning or end.
Does correlation imply causality?
1) Tertium quid - a third measurable factor of intermediate value.
2) Direction of causality - A = B but not vice versa.
Inductive reasoning
Reasoning based on probable premises.
Premise 1: I like dogs
Premise 2: I like cute animals
Conclusion: Dogs are cute.
Mill’s criteria for causality
An improvement on Hume’s:
1) Cause must precede effect.
2) Cause and effect should correlate.
3) All other explanations of cause-effect relationship rules out.
Hypothesis bias
People have a natural bias to confirm hypotheses rather than reject them.
Control condition
Condition in which cause is absent. Acts as a baseline for comparison.
Independent variable
The variable that is manipulated.
Dependent variable
The variable that is measured as a result (a.k.a. Outcome variable).
How to remove tertium quid
1) Control other factors.
2) Randomisation.
Importance of randomisation
Avoid systemic bias - to ensure an equivalent spread of attributes across groups.
How to increase confidence
1) Significance (e.g. p<0.05).
2) Randomisation.
3) Replication.
Correlational study
Naturally occurring phenomenon without interference.
Weakness = not causal.
Experimental study
Environment is manipulated in some way.
Weakness = not natural.
Why do we prefer parametric tests?
1) There is a greater variety of parametric tests so can analyse a greater variety of situations.
2) Generally better at finding experimental effects.
Self-report considerations
1) Content validity
2) Criterion validity
3) Factorial validity
Content validity
All the facets of the construct are being represented in the test.
Are all aspects of the topic assessed by the test / scale?
Criterion validity
The extent to which a measure is related to an outcome
E.g if a candidate does a test for a job interview - how valid is it as a predictor of performance in the job
Factorial validity
If one factor of the questionnaire meaningfully implies that the other factor is true and they go together.
Types of scale
1) Yes / no
2) Likert (e.g. PHQ-9)
3) Visual analog (e.g. |———x—| )
Measurement error
Difference in scores on measurement scale and level of construct we’re measuring.
Cronbach’s alpha
Splitting data in half and computing the correlation for each split to measure reliability of correlation.
N.B. Must be > 0.8 to indicate reliability.
Threats to internal validity
1) Group threats - selection process
2) Regression to mean - natural reduction to the average after initial extreme
3) Time threats - events occurring over time
4) History - unrelated events
5) Motivation - human nature
6) Instrument change - inc. calibration of equipment
7) Differential mortality - dropout effect
8) Reactivity - by measuring behaviour, it may change
Threats to external validity
1) Extraneous variables - other things that might influence result.
2) Over-use of sample groups (i.e. students, volunteers, etc.).
3) Restricted number of participants.
Quasi-experimental design
Patients allocate themselves / are allocated to the group based on pre-assessment behaviour.
Pros: ethical reasons.
Cons: other predisposing variables.
Advantages of between groups design
1) Simplicity.
2) Less chance of practice effects.
3) Sometimes hard for an individual to participate in all conditions (e.g. gender studies).
Disadvantages of between groups design
1) Expensive - time, effort, resources.
2) Less likely to detect effect of variables.
3) More confounding variables a.k.a. noise.
4) More individual differences.
Advantages of repeated measures design
1) Economy (inexpensive)
2) Sensitivity - more likely to measure what is required, reduces “noise”, and eliminates other factors.
Disadvantages of repeated measures design
1) Carry over effects from one condition to another - order effects.
2) Need for conditions to be reversible (i.e. can’t jump off a cliff in one condition).
How to eliminate order effects
1) Randomise order of conditions.
2) Counter balance order of conditions - one group does A-B and another does B-A.
Latin squares design
The order of conditions is counterbalanced so each possible order of conditions occurs once.
A B C D E B C D E A C D E A B D E A B C E A B C D
Multifactorial design
More than one independent variable.
e.g. age & gender effects
Mixed design
Where some of the independent variables are between groups and some are repeated measures.
Single subject design
A study to measure the behaviour of just one or a few people, to draw inferences from the findings.
Systematic variation
Due to the experimental manipulation in the design of the study (e.g. if participants weren’t randomised).
Randomised variation
Due to the natural existence of differences between people as uncontrolled factors.
ABA design
Baseline state is measured (state A), followed by treatment applied and behaviour is measured (state B), then treatment is removed after and baseline is measured (state A).
Treatment effects disappear / decrease when withdrawn and reappear when treatment is established.
Multiple baseline design
ABA on several participants.
If manipulation is irreversible or reversal is undesirable (e.g. can’t reignite depression), you can use AB design on a few participants instead of ABA. Each participant experiences their transition at a different time, meaning the manipulated change causes the effect rather than just variance.
AAAAABBB
AAABBBBB
Ethical considerations
1) Influence of experimenters.
2) Informed consent is a must.
3) Observational studies - OK if in public view. No spying!
4) Aim to debrief.
5) Deception has to be NECESSARY if used.
6) Participants have the right to confidentiality.
7) Protection from physical and psychological harm.
Structured interview
Quantitive research method commonly used in surveys.
Unstructured interview
Guidance to what is being researched but no prearranged questions, like in a radio interview. More of a qualitative research method.
Directional observation
Influencing the environment (e.g. in a lab) but observing natural behaviour.
Hypothesis (H1)
A proposed explanation based upon existing limited evidence as a starting point for investigation.
Null hypothesis (H0)
A default position where there is no relationship between two measured phenomena.
Remember: we cannot prove a hypothesis but we can reject a null hypothesis.
Directional (one-tailed) hypothesis
There will be a direct effect between two variables in one predicted direction.
Non-directional (two-tailed) hypothesis
There will be an effect between two variables but with no given direction.
Correlational study
A study to determine a relationship between 2 variables.
Face validity
1) Is it obvious what the test is measuring.
2) Does the test appear to assess the intended psychological construct.
Construct validity
The extent to which a test measures what it is claiming to measure
Cross-sectional design
A study to observe an effect at one given point in time.
Longitudinal study
A study to observe an effect over an extended period of time.
Effectiveness of an intervention
Is the intervention valid in a real world setting where other variables cannot be controlled
Efficacy
How an intervention performs in a controlled idealised setting
Ecological validity
Does it reflect real life
Extraneous variable
An extra variable usually unaccounted for
Predictor variable
Independent variable