Hello everyone,
I am new to FSL-VBM and I would greatly appreciate some help.
After reading the FSL-VBM manual and the papers published by Gwenaƫlle Douaud and Steve Smith I get the impression that the number of controls and patients have to be the same for the template.
In my study, the recruitment for Control to Patients have been done at a 2:3 ratio. Essentially the number of controls are always smaller. After controlling fro age and gender, I now have a total of 67 subjects in my study, of which 28 are controls and 35 are patients. I ran a preliminary analysis and found that there were significant difference in Control>Patient contrast. My questions are:
1) Given the stipulations outlined in the manual regarding a 1:1 ratio of controls and patients for constructing a template, how reliable is any analysis that uses a 2:3 or in this case a 4:5 ratio of Control to Patients?
2) In my patient group, I have two sub-groups of 21 (subA) and 14 (subB) patients respectively. Right now my GLM matrix has two groups and 3 EVs. EV1 is the mean of Controls, EV2 is the mean of Patients and EV3 is the contrast of Patient(subA)>Patient(subB). From these I have 4 contrasts EV1>EV2, EV2>EV1, EV3 and the inverse(EV3). Can I carry out a valid analysis within the constraints of FSL-VBM with this design matrix? Or do I need to create a separate patient only template if I want to investigate the differences between Patient(subA) and Patient(subB)?
3) Finally how important is gender when it comes to VBM? I have been able to keep the number of female subjects the same across all groups, however the number of male subjects differ. Is gender a serious confounder and will a difference in the gender ratio impact the outcome of VBM results in any way?
Thank you for any suggestion and help.
Best regards
Suman
|