Including covariates can in fact enhance your results. I fact if the covariates are well chosen and remove extra variance from the data, they often will. We can't answer "am I doing something wrong" without knowing what you actually did, but I would not inherently be concerned if the covariates improved my p-values.
However, consider exploring your data a bit. Reduce your threshold in the results without the covariates and see if any similar regions appear. Instead of thresholding at 0.95, try 0.8. These results are not statistically significant, but if you see a similar pattern at the lower threshold it suggests everything is OK and the covariates helped account for some otherwise unexplained variance and pushed you p-value down towards significance.
Do be careful In the analysis with covariates your contrasts are correct and you did not, for example, look at the effects of the covariate or a constant.
Best of luck,
Colin Hawco, PhD
Neuranalysis Consulting
Neuroimaging analysis and consultation
www.neuranalysis.com
[log in to unmask]
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf Of Manuel Blesa
Sent: May-16-17 9:14 AM
To: [log in to unmask]
Subject: [FSL] TBSS interpretation
Hi all,
Sorry for the naiv question, but I have some doubts about the interpretation of my results.
I run TBSS comparing two groups, and I did not find any difference. But when I adjust for some covariates, some differences appear. I was expecting in the other way around, to find some differences and after adjusment, the differences should reduce.
Am I doing something wrong? or there is a possible explanation, like the high dependence of the findings with the covariates. Thanks in advance.
Best regards,
Manuel
|