Oh, I meant to reply to an older post, but that didn't work apparently. Below the original post.
This person does a group analysis, with gender and age as covariates.You adviced to first do a DODS and look for a age-by-group interaction. If there's no interaction, he/she could continue with DOSS. My questions is (since I'm doing a similar analysis) whether it's also necesary to check for a age-by-gender interaction? Or is this not important since gender is a covariate instead of an independent variable?
Thanks!
Anita
Douglas N Greve Mon, 29 Nov 2010 07:27:30 -0800
Yes, that looks fine. The only other thing to decide is whether to use DODS (different offset different slope) or DOSS (different offset same slope). The default is DODS, but if you don't think that there is an age-by-group interaction, then you can get more power using DOSS. If you use DODS, then you should probably de-mean the ages (ie, subtract the grand mean age over all subjects from the age of each individual). doug
Roser Sala wrote: Hi,
We are doing a group analysis using qdec with two groups, and we would like to include gender and age as covariates.
Which is the best option in order to perform a simple t-test (group1>group2 and group2>group1) and including the two covariates? (we would like them to have 'zero' weighting in the contrast matrix)
We realized that if we include 'age' as a continous factor in the dialog, qdec is performing rather an interaction analysis than a covariate.
So, would it be ok to include group and gender as fixed factors and age as nuisance, and then select the output as 'Does the average thickness accounting for gender differ between group1 and group2' (age as nuisance variable)? By checking all the possibilities this seemed the best option for us, but we would like to get a confirmation that we are doing it right.
Many thanks
Roser