 |
 |
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 > ######################################################################## 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/