Good Day to you all. I am relatively new to Conn. We have submitted a longitudinal study with a small cohort (16 subjects) and analysed the effect of medication after one year using seed-based analysis. We used neurocognitive score as a covariate
We found very good results with improvement in the connectivity in several networks and there paralleled the neurocognitive improvement.
Our reviewers question
Q:1: A total of 16 seed regions were examined in the seed-to-voxel analyses, thus probable false-positives should be addressed/controlled and how conn tool box accounts the multiple comparisons.
Q: 2: Were participants instructed to keep their eyes closed or open (i.e., to focus on a white cross)? Due to the known impact of probable impact on FCs with eyes opened or closed, it is desired to include it as a covariate in second-level data analysis.
I assume that the false positives were already addressed by using FDR correction and statistical inference made at the cluster-level, utilising cluster-level threshold of p(FDR) < 0.05 after applying an uncorrected voxel-threshold of p < 0.001. The multiple comparisons in CONN are already built in the model or do we have to manually perform any steps?
when all the subjects have gone through the same protocol, is there a reason to use open eyes as covariate?
I would highly appreciate your inputs.
Siva
We found very good results with improvement in the connectivity in several networks and there paralleled the neurocognitive improvement.
Our reviewers question
Q:1: A total of 16 seed regions were examined in the seed-to-voxel analyses, thus probable false-positives should be addressed/controlled and how conn tool box accounts the multiple comparisons.
Q: 2: Were participants instructed to keep their eyes closed or open (i.e., to focus on a white cross)? Due to the known impact of probable impact on FCs with eyes opened or closed, it is desired to include it as a covariate in second-level data analysis.
I assume that the false positives were already addressed by using FDR correction and statistical inference made at the cluster-level, utilising cluster-level threshold of p(FDR) < 0.05 after applying an uncorrected voxel-threshold of p < 0.001. The multiple comparisons in CONN are already built in the model or do we have to manually perform any steps?
when all the subjects have gone through the same protocol, is there a reason to use open eyes as covariate?
I would highly appreciate your inputs.
Siva