Dear all, 

I just have a couple of, hopefully, quite quick ones. We are in the process of finalising a VBM meta-analysis with some nice results. However, we would like to include a number of covariates (notably co-morbidity and age) as covariates of both interest and no-interest respectively. What is the most appropriate way to approach this? Via the linear model comparing two groups but with no second group indicator? 

Secondly, we have carried out a couple of meta-regression analyses. Is it appropriate to use a comparisons between two groups (e.g. patients versus controls) in conjunction with a difference score between patients and controls using a measure of behaviour (e.g. depression)? For instance, cases = 40, controls = 20, so a diffeence of 20 to be included as a variable in the meta-regression?

Many thanks
Jack

Dear Jack,

an adequate way to include two covariates could be the multiple regression dialog. Note that despite you will be only interested in the first one, both variables will be treated the same.

Regarding your second question, I think it is more related to general statistics than to meta-analyses. I understand that you have two independent variables, with one being 1 for patients and 0 for controls (i.e. the group indicator), and the other being higher for patients than for controls (i.e. the depression score). Then you must take into account the same potential issues that in a standard multiple regression by two close variables, e.g. if all patients are 40 and all controls are 20 there could be colinearity, and etcetera.

Hope this helps,

Joaquim