Repeated Measures

Question 1

What are the 3 Types of within subjects or repeated measures designs?

Answer 1

1. The same behaviour for each participant is measured at different levels of the IV (e.g. concentration under different levels of noise; none, medium loud)
2. each person’s behaviour is measured repeatedly over time (e.g. ages 1, 2, 3)
3. profile analysis (several measures of the same dependant variable

Question 2

Describe the variability found in between-group designs

Answer 2

all variability within each treatment contributes to random variability between treatment means

Question 3

Describe the variability found in repeated measures designs

Answer 3

some of the variability within treatment levels can be attributed to systemic variation due to individual differences between participants

Question 4

What are the advantages of a within-subjects design?

Answer 4

• Efficient use of participants: more information is obtained from each participant.
• Variability due to individual differences is removed from the error term - increases power.

Question 5

What are the disadvantages of a within-subjects design?

Answer 5

• Practice & carryover effects (can be controlled by counterbalancing exposure to conditions)
• Differential carryover effects (can’t be controlled)

Question 6

What is a practice effect?

Answer 6

something gets better due to practice
- e.g. Accuracy rate of basketball free throws at various distances. At the later levels, may just be better because of all the practice from the previous throws

Question 7

What is a carry over effect?

Answer 7

Is something happening because it is carrying on from the previous test
- e.g. Say we rate people’s level of boredom after watching a rom com film, and also after an action film, is there a difference between the types of films of boredom? Are they more bored with time as the movies go on?

Question 8

What is the traditional approach to analysis of RM designs and what does it require?

Answer 8

• A pooled error term that separates out the effect of subjects from the main effects, and interaction
• Requires the assumption of sphericity, which is often violated
• May be more powerful for very small sample sizes (n < (a + 10)), and/or when breaches of the sphericity assumption are modest but only for the overall omnibus test.
• Unsuitable for follow-up tests

Question 9

What are the Assumptions of the Traditional Approach?

Answer 9

• Normality:
• Independence of errors:
• Sphericity (or Homogeneity of treatment-difference)

Question 10

What is the Assumption of normality?

Answer 10

Scores on the dependent variable come from populations which are normally distributed.

Question 11

What is the Assumption of independence of errors?

Answer 11

Although the scores are not independent in a within-subjects design it is assumed errors in the model are independent because individual differences between subjects have been partialled out.

Question 12

What is the Assumption of sphericity?

Answer 12

variation in scores at each treatment is similar, and also the variance in the difference in scores is similar

Question 13

What is the test for violations of the sphericity assumption?

Answer 13

Adjusted F tests to correct for violations

Question 14

How do adjusted F tests test for violations of the sphericity assumption?

Answer 14

• ε (epsilon) correction.
• e ranges from 1 (perfect sphericity) to 1/(a - 1) (extreme violation).
• When sphericity is violated, Fobt approximately follows an F distribution with adjusted df = e(a - 1), e(n - 1)(a - 1).
• The adjusted df is used to find the critical value of F to which the Fobt is compared (resulting in a larger critical value & a more conservative test).

Question 15

What adjustment is most commonly used in testing for violations of the sphericity assumption?

Answer 15

Greenhouse-Geisser
- does not alter F-obt, reduces dfs and increases F-crit - making it harder to reject null hypothesis

Question 16

Why use the traditional approach to repeated measure designs?

Answer 16

maybe more powerful test of the main effects when cell sizes are small

Question 17

What approach should always be used for follow up tests for repeated measure designs?

Answer 17

Multivariate approach

Question 18

What is the Multivariate approach?

Answer 18

• Is a multivariate extension of a dependent samples t-test
• Uses separate error terms rather than a pooled error term
• Doesn’t rely on the assumption of sphericity
• Preferred approach, though it may lack power when sample sizes are small (n < (a + 10))

Question 19

What are the 4 Multivariate Significance Tests for a Within-Subjects Factor?

Answer 19

• All F-ratios are the same unless we’re testing the interaction of a between-subject and within-subject factors
• Wilks’ Lambda
• Pillai’s Trace
• Degrees of freedom: Hypothesis (Effect): a – 1, Error: n – a – 1

Question 20

What is done to control the family-wise error in pairwise planned comparisons?

Answer 20

• bonferroni adjustment
• C = number of contrasts actually being tested

Question 21

Are there any post hoc options for repeated measures factors?

Answer 21

no
- tukey’s adjustment does not work

Question 22

For a traditional approach to a within-groups random error design what interactions exist?

Answer 22

Error of effect A
Error of effect B
Error of interaction effect (A x B)

Question 23

In a multivariate approach, what will the main effect of A & its error term will be based on?

Answer 23

the (a – 1) orthogonal difference scores of A marginal means that capture this effect.

Question 24

In a multivariate approach, what will the main effect of B & its error term will be based on?

Answer 24

the (b – 1) orthogonal difference scores of A marginal means that capture this effect.

Question 25

In a multivariate approach, what will the interaction effect of AxB & its error term will be based on?

Answer 25

the (a – 1)(b – 1) difference scores of cell means that capture the interaction.

Question 26

or within-groups random error design what is used to report on the effects?

Answer 26

Wilks

Question 27

What is a key advantage of using a paired sample t-test to follow up interaction effects?

Answer 27

can bootstrap the comparisons

Question 28

How does a traditional mixed-model approach conceptualise the design?

Answer 28

as a 3-way A x (B x S) design, with subjects as a random factor nested in the between subjects factor.
- A = between
B = Within

Question 29

What assumption applies to a traditional mixed design?

Answer 29

sphericity
- so must use epsilon (Greenhouse-Geisser) adjustments for each test involving a repeated measures factor