Should I apply 'Pool error term for RM factors: A pooled error term' ?
Hello,
I am doing a Posthoc test and considering whether I need to tick the option, Pool error term for RM factors: A pooled error term. From the suggested reference, Morey (2008), no direct explanation was found. I currently only know the option accompanied different standard errors across different conditions. How do I decide whether I take the option or not?
Thanks
Comments
Hi Pchs0114,
Pooling is generally done to create a slightly more powerful test. The "price" you pay for this is that you need to satisfy the assumption of equal variances across the levels of your factor. There are statistical tests for this (e.g., Mauchly's test of sphericity), or you can look at the observed variances and judge for yourself. It is worth noting that the RM ANOVA model also assumes these equal variances, and so if you cannot use the pooled error term for RM factors due to unequal variances, the full RM ANOVA might be misspecified.
Next year I want to revamp the ANOVA module, to allow unequal variances for the RM ANOVA model as well.
Kind regards
Johnny
Dear Johnny,
Thanks for your reply. I understand. In addition, I am looking forward to the updated module.
Regards,
Chen