PDA

View Full Version : Comparing two groups on multiple DVs



Amisa
09-25-2011, 01:08 AM
Hi all! I'm hoping for some help with reviewer comments on a paper. I'm a grad student in behavioural neuroscience, with some stats courses under my belt but nothing very helpful!

The issue is this: I have two groups (high and low on a measure), and I am comparing them on 12 different behaviours (both duration and frequency) - so 24 DVs. I am also comparing the gene expression between groups for 8 brain areas and 8 genes - so 64 DVs for this part.

The reviewers have requested a correction be made for using the multiple comparisons. However, Bonferroni (the one I know how to calculate by hand!) is WAY too conservative and ends up wiping out all my significance. I've spent the last 5 hours trying to find something that works that I can understand, and so far no luck.

Is there another correction that I can use that's more reasonable (i.e., less conservative)? I have access to both SPSS and SigmaStat, so if possible using one of these would be preferable. I can also handle fairly simple manual calculations as long as they're easy to apply to all my tests.

Thank you!!!!

trinker
09-25-2011, 07:54 AM
Do you need confidence intervals for the comparisons?

Amisa
09-25-2011, 08:29 AM
Do you need confidence intervals for the comparisons?

No - just p values or the corrected alpha level. Thanks!

trinker
09-25-2011, 01:44 PM
I was going to suggest the Holm Sidak test but then thought it may be better for you to look at the different approaches and make a decision based on your field and your data and needs.

I suggest you check out this link (LINK) (http://www.upa.pdx.edu/IOA/newsom/da1/ho_posthoc.doc) that explains different post hocs and make a decision from that. Most post hocs are fairly easy to calculate by hand (with calculator).

Amisa
09-25-2011, 05:08 PM
They actually aren't post-hoc tests, they're the main test. I am looking at 24 behaviours and want to know if my high group differs from my low group on each behaviour separately. Do they get treated like post-hocs?