 |
 |
Thank you again for your responses- I will change my analysis accordingly but I had some questions for clarification: 1a) For "group effect" I meant testing whether there is group A > group B or group A < group B. How should I approach this? I was thinking of having two between-subjects models such that 1 model would compare group A condition 1 against all conditions of group B and another model comparing group A condition 2 against all conditions of group B. So in that case would the command line look like this: palm -i A1.<ext> -i B1.<ext> -i B2.<ext> -i B3.<ext> -d design_a1vsB.csv -t contrasts_a1vsB.csv *-ise* -n 5000 -logp [blah] Then design matrix would look like the following: 1 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0 -1 0 0 1 0 0 0 0 0 0 0 -1 0 1 0 0 0 0 0 0 0 0 -1 0 1 0 0 0 0 0 1 1 1 0 1 0 0 0 0 0 -1 0 0 0 1 0 0 0 0 0 0 -1 0 0 1 0 0 0 0 0 0 0 -1 The last three columns would represent- A1-vs-B1, A1-vs-B2, and A1-vs-B3. 1b) For the one-sample t-tests command line (in this case, for Group A) you suggested- I just wanted to double-check whether I got the data correctly prepared: Command line: palm -I A1.<ext> -i A2.<ext> -d design_a1.csv -t contrasts_a1.csv -ise -n 5000 -logp [blah] For input A1.<ext> and A2.<ext> would they look like the following: A1.<ext> & A2.<ext> subject1 bh.thickness file for condition A1/A2 subject2 bh.thickness file for condition A1/A2 subject3 bh.thickness file for condition A1/A2 ...and so on Then for design_a1.csv, the file would look like this: 1 0 0 0 0 0 0 -1 1 1 0 0 0 0 0 0 1 -1 0 1 0 0 0 0 0 -1 1 0 1 0 0 0 0 0 1 -1 There would N number of columns corresponding to the number of subjects times 2 (for condition A1 and condition A2) and the last two columns would represent conditions A1 and A2. And contrasts_A1.csv would look like this: 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 Thank you very much for all the information. And I apologize again for the long email. Best, Julia Shin On Wed, Aug 10, 2022 at 1:50 PM Matthew Webster < [log in to unmask]> wrote: > Dear Julia, > If I understand your analysis correctly, you have two > groups, each having a different number of conditions, and for each group > you want to test if each of its conditions.display a positive or negative > effect. > > The simplest approach here is to split your data by both group and > condition and run a one-sample t-test ( similar to design_a1 from > Anderson’s reply in the linked thread ), the command line would then be ( > e.g. for group A ) > > palm -I A1.<ext> -i A2.<ext> -d design_a1.csv -t contrasts_a1.csv -ise -n > 5000 -logp [blah] > > With contrasts_a1.csv defining both positive and negative contrasts. > > Add -corrmod to correct across conditions, -corrcon to correct across > contrasts. > > I’m not sure how you’re defining “group effect” here, e.g. the average > across all conditions? > > Note NPC may not explicitly be needed in this case ( Anderson would be > able to clarify ). > > 2) For the second question, the same logic will apply ( it’s just a change > in storage format ) > > Hope this helps, > Kind Regards > Matthew > -------------------------------- > Dr Matthew Webster > FMRIB Centre > John Radcliffe Hospital design > University of Oxford > > > On 5 Aug 2022, at 08:37, Hye Min Shin <[log in to unmask]> wrote: > > > > 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 > > > > > ######################################################################## > > 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/