Dear Negar Chabi,
the contrast weights should be [0 1] for positive and [0 -1] for negative relationship between SPECT values and your score. To explore the results, you might want to look at the data with a less conservative threshold. Since your study includes only 21(?) subjects, it might also be worth to divide the sample by a median split and compare both groups in a two-sample-t-test without covariate.
Hope this helps
Volkmar
|