Dear @rsmith and others,
Let’s say that in a first study, I performed a one-tailed t-test with connectomestats, comparing the structural connectomes of 10 controls to 10 patients, and using NBS with cluster-forming threshold of t=2.5.
(the contrast matrix used has one line:
Now, for a second study, I have a similar sample of 10 controls and 10 patients (same disorder as in the first study), so I assume the underlying effect is the same. However, this time I want to run a two-tailed t-test instead, to see how it affects the results .
(the contrast matrix used has two line:
I know that because the p-values in tfnbs_output_wc_fwe_1mpvalue.csv are always one-tailed, I need to multiply them by two in order to get two-tailed p-values.
However, my question is regarding the cluster-forming threshold – should I keep it at t=2.5, or rather adjust it to the fact that I’m now running a two-tailed test? (i.e., t=2.5, df=18, is p = 0.0111 one-tailed; so using the same p = 0.0111 for two-tailed test, df=18, gives t=2.83)