Dear Andrew,

We ran into a question using NBS:
- We have longitudinal data (2 time points)
- 14 Subjects
- The data basis we are using is DTI streamlines (with values from 0-17'000)

We first applied a paired t-test design (as mentioned in the NBS manual) that worked perfectly.


Now we want to correlate the difference in streamlines (TP1-TP2) with the difference of a behavioral measure:

Therefore we used difference correlation matrices (TP1-TP2) and a design matrix with 3 columns: 1.) Column of ones 2.) Age 3.) Difference in a behavioral measure (TP1-TP2)

The contrast file would be [0 0 1] for testing a positive correlation between the connectivity and the behavioral measure, would that be correct? For the correlation with this difference matrices we would use a one-sample test.

For testing a negative correlation do I have to use -1 instead of 1 or do I have to multiple all correlation matrices with -1 and again use the same contrast file?

Thank you very much for your support

Hi Vivian,

Yes - you would use [0 0 1] to test for a positive correlation and [0 0 -1] to test for a negative correlation. However, you should use a "t-test", not the one-sample test.

(You could also use an F-test with any of these two contrasts, which will test for either a positive or negative correlation.)

Your design seems fine, although I think this can also be designed in terms of an interaction (I think!).

Andrew