Hi Maaike,

I know palm! :)

Please see below:

On 7 September 2016 at 09:34, maaike rive <r_maaike@hotmail.com> wrote:

Hi all,

I was adviced to use PALM for permutation of non-orthogonal surface data. However, I have some questions regarding the design matrix en contrast files to be used in PALM, but somehow I cannot subscribe to the FSL mailing list, and I was hoping one of the freesurfer users could help me out.

I have a complex design and I would like to test for interactions as well for subgroups; in addition, I have several covariates. If I use the following collumns:

A1 B1 A2 B2 cov1 cov2 cov3

I assumed I can at least test for the effects of A vs B (1 - 1 1 -1 0 0 0) and 1 vs 2 (1 1 -1 -1 0 0 0) and interactions (1 -1 -1 1 0 0 0).

Q1:

Is it true that I cannot compare subgroups, because columns denoted zero will be handeled as being a covariate? i.e. the contrast 1 -1 0 0 0 0 0 is inappropriate and I would have to build a new design matrix (A1 B1 cov1 cov2 cov3)? This is what I was told previously when using randomise for DTI data, but I'm not sure PALM works the same way.

I wouldn't say it's inappropriate, but the two models will lead to different results. Arguably the larger model will have better estimates for the coefficients for the three nuisance variables, and if the variances are the same for all subjects (as assumed anyway), the variance estimates are also better. This means that in the larger model you would use the information from the other subjects that are not part of the contrast.

Q2:

What I also understood from randomise is that you have to define both 1 -1 and -1 1; there is no option for a 2tailed t-test or F test. Is that correct and does it also apply to PALM?

In randomise F-tests can be used to test the two tails. In PALM, the F-tests only work for contrasts that have rank > 1, otherwise a two-tailed test can be done with the option -twotail. Another possibility in PALM is to, instead of using F-tests or -twotail, use the option -corrcon, that will correct across all contrasts.

All the best,

Anderson

Thanks,

Maaike

_______________________________________________
Freesurfer mailing list
Freesurfer@nmr.mgh.harvard.edu
https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer

The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.