Experimental research designs

Question 1

Experimental versus quasi experimental

Answer 1

Quasi experimental variable- no random assignment, likely because this is possible/ethical. Quasi independent variables include age, gender, ethnicity.
Experimental approach- random assignment to conditions.

Question 2

Confounding variables

Answer 2

o Person confound: individual differences that co-vary with experimental conditions (especially could be the case with one quasi experimental variables like gender and manipulating the other variable)
o Procedural confound: manipulating factors other than intended IV
o Operational confound: same effect, different ‘cause’

Question 3

Solutions to reduce confounding variables

Answer 3

• Proper random assignment procedure:
o Large PPs help reduce person confound
• Take care with manipulation:
o If it is too obvious, will result in demand characteristics
o Too socially undesirable?
• Measure potential covariates:
o Check questions
o Statistical control
• Minimise noise, have as much control over this as possible:
o Lab vs. field
o Reduce bias, can use double blind procedures
• Be systematic:
o Record keeping
o Automating instructions
o Noting possible influences on findings, etc.
• Artificiality (biggest disadvantage):
o Trade-off between internal and external validity

Question 4

Disadvantages for between subjects designs

Answer 4

In a between-subjects design, error variance is the variability in the scores within each condition. As the
scores within each condition become more spread out, error variance increases. As error variance
increases, it becomes harder to detect whether an effect exists. Between subjects designs tend to have higher error variance than within-subjects designs.
Also in between subjects, groups may differ on a variable just be chance, rather than due to an manipulation of the IV. It is therefore important in between subjects designs to ensure that a pre-test is carried out to make sure that there is no pre-existing difference between groups on the IV.
Also use matching, select pairs of ps who are similar to one another. Use this in conjunction with random assignment –> high internal validity

Question 5

Post-test only comparisons

Answer 5

o Compare to control, and compare to another level of IV
o Measurement of all groups one at the same time as it reduces threats through history or maturation?
o Weaknesses:
 No pre-test, so can’t make statement about actual change or effect
 Not sure which group actually changed
 If the findings differed for 2 groups, how many changed and which groups changed?

Question 6

Pre-test post-test control group design.

Answer 6

o Comparison of 2 or more groups with random assignment
o Pre-test measures taken for all groups (mixed)
o Compare to control and to another level of treatment
o Can make better claim about change in scores
o Weaknesses:
 Measuring twice may influence responses
 Treatment effect may only occur with pre-test, in which both internal and external validity is limited

Question 7

Solomon 4 group design

Answer 7

o Random assignment to one of four groups:
 Group 1: pre-test/treatment/post-test
 Group 2: pre-test/post-test
 Group 3: treatment/post-test
 Group 4: post-test only
o Groups 1 and 2: basic pre-test/post-test element
o Groups 3 and 4: evaluates treatment in absence of pre-test
o Expensive and time consuming
o Effective for reducing many internal validity threats

Essentially combines post-test only with pre-test post-test design

Question 8

Within subjects design strengths and weaknesses

Answer 8

Repeated measures
• PPs do all conditions
• Avoids problems with PP variability
• More power where PP effects on DV are substantial
• Fewer PPs required
• Weaknesses:
o Effects could be due to maturation, history, testing (reduces with practice/breaks)
o May be more attrition
o Order effect: try to minimise with counter balancing
 Counter-balancing works if carry-over effects induced by different orders are approximately equal
 Can test for order effects by making them an IV, and also compare with between-PPs designs
 Counter-balancing is not necessary if treatments/ IV levels can be randomly intermixed
 Intermixing can be the only means of controlling for confounding strategic effects, e.g. manipulation of any factor impacting task difficulty is confounded with PP effort when manipulation is blocked

Question 9

Single N designs

Answer 9

• Rare
• Case studies, or single group is used
• Can be experimental repeated-measures design
• Measure baseline to post-treatment (A-B)
• Or measure baseline, post-treatment, withdrawal of treatment (A-B-A)
• Useful for rare cases (e.g. testing new drugs) or where extensive training is required

Question 10

Factorial design

Answer 10

• More than 1 IV – effect of each IV on DV and combined effects of IVs on DV
• Simple main effect: effect of one IV at the specific level of the second IV
• Main effect: average influence of one IV on DV
• Interaction: levels of one variable have different effects depending on level of other variable
• The more factors, the more complex it becomes