Factorial ANOVA

Question 1

What is meant by a 3 x 4 design?

Answer 1

2 IVs, one with 3 levels and one with 4 levels

Question 2

ANOVA is very robust to breaches of the normality assumption provided that…

Answer 2

• similar number of participants in each cell
• there are 10+ participants in each cell
• even if skewness or kurtosis is present the effects on ANOVA are small

Question 3

What does it mean if there is a significant interaction effect?

Answer 3

the effects of one facter will be different depending on the other factor
- need to examine effects of one factor separately at each level of the other factor
- test for simple main effects

Question 4

What is the equation for partial eta-squared?

Answer 4

SSeffect/SStotal

Question 5

What is partial eta-squared used for?

Answer 5

give a better comparison of relative effect sizes of our IVs within the study
- it gives us the effect size for the IV that is not contaminated with any additional effects of the other IV on DV

Question 6

What is the effect size used for all F ratios?

Answer 6

eta-squared

Question 7

What effect size is used for follow up tests?

Answer 7

Cohen’s d

Question 8

What is power effected by?

Answer 8

1. Size of treatment effect
2. Error variance
3. Sample Size
4. The alpha level

Question 9

How do you increase power?

Answer 9

• Maximising the effect of the IV
• Reduce error variance
• Increase sample size
• Relaxing the alpha-level (e.g. .10)

Question 10

How can power be reduced?

Answer 10

• using a weak IV manipulation
• random variability in the data
• using too small a sample size
• reducing the alpha level (e.g. .01)

Question 11

What do each of the following numbers represent?
F(2,54) = 26.67, p < .001

Answer 11

2 = df (of the IV tested)
54 = df of the error
26.67 = F ratio
.001 = Sig.

Question 12

How do you follow up main effects?

Answer 12

1. EM Means
   - compare main effects and get pairwise comparisons
2. Post Hoc test (multiple comparisons)
3. Conducting Planned comparisions

Question 13

If NO Interaction Effect follow up Main Effects by…

Answer 13

If 2 levels no more follow up needed
If 3 or more levels follow up with planned or post hoc comparisons just like 1 way ANOVA
- But uses pooled error term from factorial ANOVA
- Consider Family wise Type I error rate v Type II error and justify your decision for the alpha level used (remember post hoc’s automatically adjust for all possible comparisons)

Question 14

What is Bootstrapping?

Answer 14

randomly re-sampling from within our sample data

Question 15

What is the simple bootstrap method?

Answer 15

When randomly selected, the participants in group 1 for the original data will remain in group 1 the data will be kept together

Question 16

What is the stratified bootstrap method?

Answer 16

where you have 2 or more conditions, and you randomly select cases from within each condition. Within each condition sample, you take resamples of size n (little n) with replacement.

Question 17

When bootstrapping what do you do to the CI intervals?

Answer 17

adjust them to 97.5% Bias corrected accelerated (Bca)
- e.g. Bca 97.5% CI [1.1, 2.2]