df for Random Effects ANOVA
Is it possible to get a df that's not an even number when doing a random effects ANOVA? For my random effects homework, I'm getting non-even number errof df, and a higher F but also higher p than for the fixed effects ANOVA. How is this possible?
posted by Unknown @ 11:05 AM
2 Comments:
At 2:44 PM,
Mari Clements said…
In the examples in class today and yesterday, we had exactly this situation...
At 5:22 PM,
Mari Clements said…
Remember also that an F is uninterpreatble without its associated degrees of freedom. That is, the lowered df for the error term for main effects in a random effects design hurts your power in two ways: by its effect on the MS of the error term and by its effect on the critical value for F. An F that is significant with 1 and 46 df will not necessarily be significant with 1 and less than 46 df.
Post a Comment
<< Home