Dear Alfonso,

>Hello,
> We have got some differences on MTLobe when comparing MCI and AD patient
>groups, besides controls, on an encoding paradigm based on voxel-by-voxel
>statistics for fMRI. Volumetry of hippocampus also showed significant
>differences among all groups. A reviewer suggested us to introduce volumetric
>measures as a covariate to study structure-function relationship. However, we
>have some problems to understand how suitable or how to interpret our results.
>Why? Because when we introduce left+right hipocampal volume as a single
>covariate (in a contrast defined as nuisance [0 1 -1 0 0]at ANCOVA RFX level),
>differences among all groups (mainly between MCI and AD groups) increased
>spectacularly. Is it correct to introduce that covariate? And, then, may we consider that once
>variance on fMRI data explaned by hippocampus volume fMRI increased differences
>may be attributed to functional activity?
>  
>
I think your interpretation is correct. In the model you set, you had 3 
variables (gp1, 2, 3) that explained the variance around mean. In the 
new model, you have 4 variables: 3 variables correspond to groups and 
one to volume. The contrast [01-100] compare the activity of groups 2 
and 3 once the variance related to the hippocampus volume is regressed 
out. In the first model, this part of variance wasn't explain and was 
thus in the error term. With this covariable (new model), the error term 
is smaller and your statistic increases :-) . In conclusion, you have a 
true functional difference that cannot be assigned to a volumetric 
difference.

>On the other hand, how may I introduce two covariates in an ANCOVA? Since we
>were controlling for sex differences previously and we would like to control
>for both covariates (sex and volumetry).
>  
>
well, you can either modify the spm code (see below) or simply use a 
multiple regression model with 5 covariables (gp1, gp2, gp3, cov1, cov2) 
and perform a gp2 vs gp3 contrast.

Best
- Cyril

For multiple cov in ANCOVA, modify spm_spm_ui.m
-- go to ANCOVA part --
-- find the line --
'nC',[0,1],'iCC',{{8,1}},'iCFI',{{1,1}},...
-- and substitue by --
'nC',[0,Inf],'iCC',{{8,1}},'iCFI',{{1,1}},...