thanks jesper, I do actually need good luck.
i understand your first scenario (age and disease duration VS MR-metric) and your suggestion (recruit correlation-breaker subjects is smart indeed).
regarding the second scenario...
>>>you have performed a whole battery of psychometric tests on your subjects and it turns out you find no correlations between your MR-metric and any of the tests. In this case it is quite possible that the problem is that many of those tests asses the same or similar functions and that they are therefore highly correlated.
are you meaning, as i was asking, that each single test correlate but when all these test are putted in a same GLM matrix the effect disappear in all the contrasts ???
well, i understand that if n test are highly correlated, "none of them prevails in correlating" with my MR-metric, everyone compete to explain part of the variance and they all "lose" not reaching the significancy, but this should be different from "none of them anymore correlate" with the MR-metric....especially from a clinical point of view.
how can i manage it ? . do N single test and report voxels with <0.05/N ???
suppose that i measure 5 score, single analysis correlate with 4, when i put those four on a single GLM, nothing comes out. if i put only 2 (A,B), i get something... now i could have really measured that two and report them, but i could also have measured 5, and told the i measured exactly those 2..or i could have really measured one of them (A) but also one different (C) and nothing comes out...so i lost the one really(?) correlating (A).
with some of confusion, i'm trying to say that it should not be a matter of being luck or honest. single-score analysis should still maintain a significance. there should be a way to report them.
supposing these relations:
more axonal damage (less FA) may be related to more neuronal damage (less NAA/Cr)
more axonal damage (less FA) may be related to an higher disease severity
since the the latter should be positively related also to more neuronal damage, if i look for MR-metric relation with both variables i may lost lot of effects, why this should happen?
thanks for your time
regards
Alberto
|