Title: Research methods in clinical psychology: An introduction for students and practitioners Chris Barker
1Research methods in clinical psychologyAn
introduction for students and practitionersChris
Barker, Nancy Pistrang, and Robert Elliott
- CHAPTER 8
- Foundations of design
2Design overview
- Types of designs
- Validity analysis
- Descriptive and correlational designs
- Correlation and causation
- Quasi-experimental designs
- Randomised experimental designs
- Control and comparison groups
3Classification of designs
- Non-experimental designs
- Descriptive
- Correlational
- Experimental designs
- Non-randomised (quasi-experimental)
- Randomised
4Causal relationships
Central principle Correlation does not mean
causation
Suppose that A and B are correlated what might
be the causal relationship between them?
ctd./
5Causal relationships 2
Simple causation
B
A
B
A
A
B
Third variable
C
ctd./
6Causal relationships 3mediating and moderating
variables
Mediating variable
A
D
B
Moderating variable
A
B
E
(Baron Kenny, 1986)
7Validity analysis
- Statistical conclusion
- Internal
- Construct
- External
- (from Cook and Campbell, 1979)
8Statistical conclusion validity
- Is the study sensitive enough?
- statistical power
- design quality
- Do the variables covary?
- If so, how strongly?
9Internal validity
- If two variables do covary, is there a causal
relationship between them?
10Construct validity
- Do the outcome variables (and also the
experimental intervention) represent the
underlying constructs that they are supposed to?
11External validity
- Do the results of the study generalise (e.g.,
across settings or participants)?
12Some non-randomised designs(Cook Campbell,
1979)
- (Notation
- X intervention
- O observation)
- One-group posttest-only design
- X 0
- One group pretest-posttest design
- O1 X O2
13Some non-randomised designs (cont.)
- Nonequivalent groups posttest-only design
- NR X O
- NR O
- Nonequivalent groups pretest-posttest design
- NR O X O
- NR O (Y) O
14Some threats to internal validity
- Endogenous change
- Maturational trends
- Reactivity of measurement
- Secular drift
- Interfering events
- Regression to the mean
- Selection effects
15Some threats to the construct validity of the
intervention
- Confounding variables
- non-specific factors
- Expectancy effects
- placebo effects
- Hawthorne effect
16Randomised designs
- Essential feature randomised assignment to
experimental groups (conditions) - This controls for the internal validity threat of
selection effects, which is a problem with
quasi-experimental (non-randomised) designs
17RCTs
- RCT is a common abbreviation for either
- Randomised Controlled Trial
- or
- Randomised Clinical Trial
- Efficacy (as opposed to effectiveness) research
18Example
- Randomised groups pretest-posttest design
- R O X O
- R O (Y) O
- (see Cook Campbell, 1979)
19Randomised designs terminology
- The dependent variable(s) are the outcome
measures - The independent variable(s) are the experimental
conditions - The statistical method is analysis of variance
(ANOVA)
20Factors in randomised designs
- Independent variables are arranged as factors
- each has two or more levels
- multifactorial designs
- within-group (repeated measure) factors
- between-group factors (experimental conditions)
- blocking factors (participant differences)
21Some types of control group
- No-treatment controls
- treatment better than nothing
- Wait-list controls
- control for expectation of benefit
- Placebo controls
- credible but inert treatment
- double- and triple-blind
- Comparative treatment groups
- Alternative treatment
- Treatment as usual
- Dismantling studies
22Practical limitations of randomisation
- Non-equivalence of groups
- Attrition
- intent to treat analysis
- Leakage
- Non-cooperation of staff
- Costly and time consuming
- Cant study negative events
- Ethical issues
23Ethical issues in RCTs
- No treatment/placebo controls
- Wait-list controls
- Random assignment v. choice
- Specified treatments v. clinical judgements
- Decision on patient inclusion
- Referrals at termination
24Some good experimental design features (1)
- Patient homogeneity
- Randomised assignment
- Groups similar after randomisation
- Specific interventions
- manualisation
- Appropriate control groups
- Groups treated equivalently except for
intervention - ctd./
25Good design features (2)
- Low attrition
- Patients, clinicians and raters blind
- Follow-up after termination (e.g., 6 months, one
year) - Independent replication