I'm wanting to conduct a 2 (phase: P1, P2) x 2 (group = G1, G2) mixed model ANOVA, whereby phase is the continuous within subjects variable and group is the between subjects variable.

The phase variables are taken from response-time data (e.g., mean ms at P1 = 960) and are both normally distributed (p > .05). My group sizes are relatively equal (G1 = 46, G2 = 58).

I found a significant main effect of Phase, F (1, 102) = 1049.75, p <.001, suggesting that participants were faster at one phase than the other, but no main effect of group (p >.05), nor a significant interaction term (p > .05). These findings I'm happy with. However, the data violated Levene's assumption of homogeneity of variance for Phase 2, but not for Phase 1. I log-transformed the data for both phases (even though they are normally distributed), but the results and violations are the same.

1) What should I do about this violation? Should I just run a one-way ANOVA with Welch's correction to account for heterogeneity and determine whether there is a main effect of group on phase scores? Should I then run a t-test to explore the main effect of phase? Or is a factorial ANOVA robust against slight violations of homogeneity (p = .02)?

2) Is the violation to homogeneity of variance the reason for my large f statistic? It seems too big of a statistic.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.