Hi Doug,<br><br>using mri_glmfit with DODS and mri_glmfit-sim with a fsgd-file and contrast (attached) that looks like (2 discrete factors: 1 with 2 levels (diagnosis) and 1 with 4 levels (Site) and 1 continuous covariate (age)):<br>
GroupDescriptorFile 1  <br>
Title AGE-subj  <br>Class PatSite1  <br>Class PatSite2  <br>Class PatSite3  <br>Class PatSite4  <br>Class ConSite1  <br>Class ConSite2  <br>Class ConSite3  <br>Class ConSite4  <br>variables Age<br>Input $FSGDID PatSite1 12.8<br>

..<br>Input $FSGDID PatSite2 9.3<br>..<br>Input $FSGDID PatSite3 15.1<br>..<br>Input $FSGDID PatSite4 10<br>..<br>Input $FSGDID ConSite1 7.6<br>..<br>Input $FSGDID ConSite2 8.9<br>..<br>Input $FSGDID ConSite3 10.2<br>..<br>

Input $FSGDID ConSite4 14<br>..<br><br>mri_glmfit --y lh.justmen.NOTmatched.CT.fwhm10.mgh --fsgd fsgd_justmen_NOTmatched_only_age.txt --C contrast_justmen_DODS_GxA.mtx --surf fsaverage lh --cortex --glmdir lh.justmen.NOTmatched.all.GRxAge.CT.fwhm10<br>

<br>for the contrast attached (which i thought is the interaction between diagnosis and age, independent of &#39;Site&#39;) i found 3 clusters using mri_glmfit-sim (attached). i saved these clusters as labels and transformed them to native space (cluster appeared in the correct location in native space):<br>

mri_label2label --srclabel cluster-001.label --srcsubject fsaverage --trgsubject $subject --trglabel $subject_cluster-001.label --regmethod surface --hemi lh<br><br>and extracted mean CT from the cluster in native space:<br>

mris_anatomical_stats -l  $subject_cluster-001.label -t lh.thickness -b -f $subject_cluster-001.stats $subject<br><br>i plotted mean thickness for each cluster against age but did *not* see an interaction between groups. even though this is mean thickness for the cluster on non-smoothed native space thickness i would have expected to see some interaction but the slopes for both groups are almost parallel. any thoughts? i can provide more info if needed.<br>

-joost<br>