Dear All.
We are investigating a rare genetic disease that causes mental decline and slowly progressive brain atrophy. Our study group consists of 12 patients and 36 age-and sex-matched control subjects imaged with 3T. In other words, there are 12 “matching groups”, each consisting of the patient and the three matched controls. At the moment we are performing VBM with SPM8 and VBM8 toolbox, but we are unsure how to construct the design matrix with the three matched controls per patient, and would truly appreciate your comments on this.
Our initial idea was to calculate an “average control” from the three matching controls and use a design matrix which calculates paired t-test between the patients and averaged controls. Then we thought, it would be better to take the multiple controls into account in the model instead of averaging the matching controls. From the literature, we found no examples on how to build design matrix to handle several matched controls so we are a bit unsure of our approach. We used option ‘Full factorial analysis” to construct the following design: First column of the design matrix specifies the patients and second column specifies the controls. Third column specifies the first matching group (i.e. patient #1 and the 3 matched controls #1a, #1b, #1c), fourth column the second matching group (patient #2 and controls #2a, #2b, #2c) etc. Columns from 3 to 14 would therefore model the “matching group effect”. In total, there would be 2+12=14 columns in the design matrix and the contrasts to study differences between the patients and controls would be [1 -1 0 0 ...] or [-1 1 0 0 0 ..].
In practice, we used “Full factorial model” of SPM8 to create the design matrix. The 2-level factor was used to distinguish patients and controls. Then, we used 12 covariates to specify the 12 matching groups. We can send you the resulting design matrix if necessary.
Can you see any statistical caveats in performing the analysis in the presented way or would you suggest some other approach?
Thank you in advance!
Kind regards,
Anna
|