4: REPEATED MEASURES ONE-WAY ANOVA

Question 1

RM designs (one-way ANOVA): what contributes to variance

Answer 1

between IV:
- manipulation of IV (treatment effects)
- experimental error (random /constant)
RM designs: variance between IV levels due to individual differences is absent

within IV:
- experimental error (random error)

RM designs: we remove the variance due to individual differences from the variance within IV levels (still constitutes a part of variance just neither within/between)

Question 2

t/F ratio (RM)

Answer 2

t/F = variance between IV levels / variance within IV levels (excluding individual diffs)

• between - includes variance ‘caused’ by our manipulation of the IV and error variance
• within - includes only error variance

Question 3

F ratio : RM

Answer 3

F = variance between IV levels / ( variance within IV levels - individ diffs)
F = MSm/MSr

• F close to 0 - small variance (relative)
• F further from 0 - large variance (relative)

Question 4

assumptions of repeated measures 1-way ANOVA

Answer 4

• normality: the distribution of difference scores under each IV level pair should be normally distributed
• sphericity (homogeneity of covariance): the variance in difference scores under each IV level pair should be reasonably equivalent (Mauchlys assesses this, greenhouse corrects for this)
• equivalent sample size: sample size under each level of the IV should be roughly equal

if data violates, non parametric - friedman test

Question 5

Mauchly’s statistic

Answer 5

assesses sphericity (homogeneity of variances)

therefore, if P < (or equal) .05 we reject null hypothesis (i.e. heterogeneity)

SPSS:
- if Mauchly’s is not significant use sphericity assumed
- if Mauchly’s is significant use greenhouse-geisser

Question 6

df for RM 1-way ANOVA

Answer 6

need to calculate df for our estimates of :

• between IV levels: DFm = k-1
• within IV levels: DFr = DFm x (n -1)

Question 7

advantages of repeated measures designs

Answer 7

• recruitment: needs fewer participants to gain the same number of measurements
• error variance (within IV levels) is reduced (remove individual diff variance)
• more power with same number of participants (easier to find significant difference (avoid type 2 error, resulting F/t value is larger)

Question 8

disadvantages of RM designs

Answer 8

order effects: practice effects, fatigue effects, sensitisaion, carry-over effects
- use counterbalancing

alternatives where counterbalancing not possible: practice, fatigue, sensitisaion, carry-over effects