Thank you for the response Matthew.

So if I still want to NPC by still correct over both hemispheres, can I merge lh and rh into bh using palm_hemimerge and then split them later using palm_hemisplit as discussed in this thread: https://jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;42baf4dd.1608

I also have a couple more questions if that is ok:

1) After going through the repeated measures anova tutorial, I wanted to try to apply PALM NPC in my analysis. I have 2 groups with group A having 2 conditions (A1 and A2) and another group B with 3 conditions (B1,B2,B3). I ran all 5 conditions on each subject. At first I just wanted to analyze them by having 10 t-tests and 2 F-contrasts (looking for the effect of group A and the effect of group B) then correcting for multiple comparisons using PALM/NPC. Would this be a good choice for my analysis?

I also considered using NPC to combine within-subject and between-subject models after reading this thread: https://jiscmail.ac.uk/cgi-bin/wa-jisc.exe?A2=ind1801&L=FSL&P=R60203&K=2

I thought this would be more appropriate for my analysis but with how my conditions are set up, it looks unfeasible since I do not have blocks with the same size for whole-block EB.

My initial idea for within block and whole block EB is the following:

1 -subject 1 A1

1 -subject 1 A2

1 -subject 2 A1

1 -subject 2 A2

2 -subject 1 B1

2 -subject 1 B2

2 -subject 1 B3

2 -subject 2 B1

2 -subject 2 B2

2 -subject 2 B3

So would it be better to just stick with 10 t-tests and 2 F-tests using NPC?

2) If I can go ahead with this joint NPC analysis, can I also use this on subcortical data?

Sorry for these additional questions- your responses were very helpful to understand PALM/NPC and how I should analyze my study.

Thank you in advance,

Julia Shin

On Tue, Aug 2, 2022 at 3:37 AM Matthew Webster <[log in to unmask]> wrote:

Hello Julia,
Note NPC should not be used here as it requires all inputs to be aligned - and by definition the left and right hemispheres are un-aligned. The -corrcon and -corrmod options can still be used to correct over both hemispheres ( “modalities” ) and contrasts though.

Kind Regards
Matthew

--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
On 29 Jul 2022, at 14:00, Hye Min Shin <[log in to unmask]> wrote:
No problem- I was able to successfully run the F-test using PALM with npc.

So if I also wanted to correct for both right and left hemispheres while applying npc to correct for contrasts can I add "-corrmod" and "-npccon" to my palm command? For a repeated measures design like the one is FreeSurfer tutorial, would it be ok to only apply npc to correct for contrasts? Or do I also have to add "-corrcon" to correct for contrasts AND add "-npccon".

The command would look like the following (as also noted in this tread- https://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=FSL;42baf4dd.1608):

palm -i lh.thickness.sm15.mgh -i rh.thickness.sm15.mgh -m lh/mask.mgh -m rh/mask.mgh -d design.mat -t design.con -f design.fts

-n 5000 -T -npc -corrmod -npccon -logp -approx tail -nouncorrected

Best,

Julia Shin
On Wed, Jul 27, 2022 at 9:28 PM Matthew Webster <[log in to unmask]> wrote:
Hello Julia,
Ah - my mistake! The F-test is on the _first_ two t-contrasts so the F-contrast should be 1 1 0 0 as you say.

Kind Regards
Matthew

--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford
On 27 Jul 2022, at 13:18, Hye Min Shin <[log in to unmask]> wrote:
I apologize for sending this email again- I originally responded to Matthew's email from JISCM@il site and since I can't see the message thread I am posting my response through my school webmail. 
Thank you very much for your reply Matthew- I am currently trying to run PALM again with your suggested F-test matrix of 0 0 1 1.

My question is what does the F-test matrix 0 0 1 1 mean here. 

I thought the columns of the F-test matrix would correspond to the rows of the t-test matrix. So if t-test matrix has row 1 with tp1-vs-tp2, row 2 with tp1-vs-tp3 then the F-test matrix would be 1 1 0 0 so that row 3 and row 4 of the t-test matrix which are tp2-vs-tp3 and subject mean would not be chosen. 

Also in my palm command I gave the option -cmcx and I was wondering if this would be an appropriate flag to use instead of -npccon as suggested in PALM/JointInference user guide. At first I did give -npccon option but I got an error so I switched to -cmcx as suggested in the message. 

Best,
Julia Shin
On Wed, Jul 27, 2022 at 8:15 PM Matthew Webster <[log in to unmask]> wrote:

Hello Julia,
You only need to include ( any ) 2 of the 3 tp_x-vs-tp_y t-contrasts as the 3rd difference is implicitly included in the test - e.g. if there is no difference between tp1,tp2 and tp1,tp3 then by definition there is no difference between tp2,tp3.

With reference to the Freesurfer link in your email, this would be equivalent to "Contrast 4 (tp-effect.mtx)" which in FSL/PALM terms would be F-contrast 0 0 1 1. The F-contrast in your email ( 1 1 1 1 ) as well as being rank-deficient, also includes the subject-mean t-contrast, which may or may not be what is intended.

Kind Regards
Matthew

--------------------------------
Dr Matthew Webster
FMRIB Centre
John Radcliffe Hospital
University of Oxford

> On 27 Jul 2022, at 11:01, Hye Min Shin <[log in to unmask]> wrote:
>
> Hello experts,
>
> I am trying to run FSL PALM specifically NPC using the example from the freesurfer repeated measures anova tutorial page: https://surfer.nmr.mgh.harvard.edu/fswiki/RepeatedMeasuresAnova#Contrast4.28tp-effect.mtx.29
>
> This freesurfer tutorial has 3 subjects with 3 timepoints with the following design matrix (column1-3 are subjects 1-3 and column 4 is tp1-vs-tp3 and column 5 is tp1-vs-tp3):
>
> 1 0 0 1 1
> 1 0 0 -1 0
> 1 0 0 0 -1
> 0 1 0 1 1
> 0 1 0 -1 0
> 0 1 0 0 -1
> 0 0 1 1 1
> 0 0 1 -1 0
> 0 0 1 0 -1
>
> It also has 5 contrasts, 4 t-tests and 1 F-test.
>
> The design matrix for the t-tests is the following:
>
> 0 0 0 1 0 (tp1-vs-tp2)
> 0 0 0 0 1 (tp1-vs-tp3)
> 0 0 0 -1 1 (tp2-vs-tp3)
> 1 1 1 0 0 (is the subject mean == 0? )
>
> The design matrix for the F-test is the following:
> 1 1 1 1
>
> I wanted to run the F-test to see if there is a group effect or specifically in this case the effect of timepoint. However, whenever PALM tries to run the F-test using NPC it keeps saying the rank is deficient.
>
> I ran PALM using npc with the following command:
>
> palm -i ./lh.thickness.sm15.mgh -m ./a_glmfit-lh/mask.mgh -d ./a_glmfit-lh/palm-onetail-1.3-lh/design.mat -t ./a_glmfit-lh/palm-onetail-1.3-lh/design.con -f ./adni_glmfit-lh/palm-onetail-1.3-lh/design.fts -n 5000 -npc -cmcx -o ./a_glmfit-lh/palm_mv-lh -logp
>
>
> Is there something wrong with my F-test matrix or do I not need to run the F-test if I was able to run the t-tests on paired comparisons?
>
> Thank you in advance for your help
>
> Best,
> Julia Shin
>
> ########################################################################
>
> To unsubscribe from the FSL list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
>
> This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/

########################################################################

To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1

This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1