Dear SPMers,
I am performing a 'covariates only' statistical design using clinical diagnosis as a nuisance variable.
I assigned an arbitrary value to each group diagnosis (e.g. 1 = Alzheimer's Diseases, 2 = Parkinson's disease, 3 = cognitively unimpaired subjects).
I re-run the same analysis changing the values associated with each diagnosis diseases (e.g. 2 = Alzheimer's Diseases, 3 = Parkinson's disease, 1 = cognitively unimpaired subjects).
Despite showing overlapping maps, the results (to my surprise) showed:
1) differences in significance values (e.g.T=4.12 in the most significant voxel in the first analysis, T = 3.84 in the same voxel in the second analysis);
2) Slightly different peak location within significant clusters.
How could I interpret these differences of significance?
Thanks in advance for your help,
Simona
|