Reply With QuoteUPDATE: so i've just read in one stats book that omega squared cannot be calculated with unequal-sized samples - so you can't get effect sizes. Is there a way around this?
Hi all,
I was wondering if anyone could help me out. So i'm trying to calculate effect sizes for some data following an ANOVA. However, i need to first compute a variance component for the main effects and the interaction in order tio calculate omega squared. the denominator in the equation is "nab" where
n=number of people per condition,
a = levles in 1st IV
b = levels in 2nd IV.
So, heres the problem: theres differant amount of people in each condition (e.g. condition is gender andtheres 15 males and 32 females). What do i do!!??
Any help would be great, OR if people know other ways to get effect sizes on data without doing this, this would be great!!
UPDATE: so i've just read in one stats book that omega squared cannot be calculated with unequal-sized samples - so you can't get effect sizes. Is there a way around this?
Omega squared is not effect size. It is the amount of variance explained. Do you need effect size or to address the amount of variance explained?
If it is the former then I dont think unequal sample size means you can not calculate effect size. If it is the later you can try eta squared which is less optimal than omega squared. It can I believe be calculated with unequal group size although bias will occur in some cases (eta squared is generally assumed to be biased but still commonly used).
http://jom.sagepub.com/content/24/2/157.abstract
"Facts are stubborn things, but statistics are more pliable." Mark Twain
Posting Permissions
|
|