Onesided or twosided test in the new version of fixelcfestats

1

Dear experts,

I noticed that the -negtive option is nolonger available in the new released version in fixelcfestats. I’m confused about this. For example, if I want to do a one sample ttest, the output files like fwe_1mpvalue.mif and uncorrected_pvalue.mif, are they onesided or twosided results? Also, I want to make sure that the p values in these two files in fact are the values of 1-p, right?

Thanks for your time,
Chunxiang

2

I think the answer to that will be the same as for mrclusterstats – see this thread here…

3

Always one-sided for t-tests. The old -negative option was a hack just to expedite testing opposing contrasts within the constraints of the old GLM code and with reduced computational time (since prior to 3.0.0, if it weren’t for the -negative option, you would need to generate the fixel-fixel connectivity matrix independently each time). The 3.0.0 code allows an arbitrary number of t-tests to be defined in the contrast matrix, one line per test, which in terms of capabilities is a superset of just testing the opposite contrast.

Also, I want to make sure that the p values in these two files in fact are the values of 1-p, right?

Correct. I changed the name of the relevant file from “fwe_pvalue.mif” to “fwe_1mpvalue.mif” in the hope of better communicating this, but I agree that it’s far from obvious. I would also note that given the various mrview toolbars now support upper & lower thresholds and the inversion of colour maps, some of the justification for writing p-value data in this form has eroded…

4

Thank you for your response @jdtournier @rsmith. I also came across this post: FBA design / contrast matrices for three groups, which answered most of my questions with regard to the statistical analysis. Hope more people who have similar confusions find it helpful.

Best,
Chunxiang