 |
 |
Hello Stefan, can you check that your randomise call refers to the correct con and mat files, respectively? Perhaps post the command line you are using. Cheers, Cornelius On Thu, Apr 18, 2013 at 9:52 AM, Lenzen, Stefan <[log in to unmask]>wrote: > Hi, > > I put brain volume, motion parameter, age and education as covariates into > the analysis. However, I was not interested in the effects of these > variables between the groups (so marked them as "0" in the contrast file). > I just wanted to know the difference between the two groups, when the > described covariates above are controled for). And it doesn't matter > whether these covariates were put into the analysis or not (so I even tried > to delete all the columns encoding the covariates as well in the .mat as in > the .con files, so only the EVs encoding the groups remained, but still the > results were the same). So none of these change anything in the output > (absolutely identical). > > Thanks and kind regards, > > Stefan Lenzen > ------------------------------ > *Von:* FSL - FMRIB's Software Library [[log in to unmask]]" im Auftrag > von "Himanshu Joshi [[log in to unmask]] > *Gesendet:* Donnerstag, 18. April 2013 08:23 > *An:* [log in to unmask] > *Betreff:* Re: [FSL] Identical Output from dual_regression despite > altered .con and .mat files > > Hello Stefen, > Do let us know what contrasts are you putting in your design matrix for > the explanatory variables having identical results when these ev's are > present or absent . > > Thanks > > > On Wed, Apr 17, 2013 at 10:10 PM, Stefan Lenzen <[log in to unmask]>wrote: > >> Dear FSL experts, >> >> when performing a dual regression analysis of my data, I encountered a >> problem with the design matrix. I'm investigating the differences in >> resting state of 2 groups, while correcting for several covariates (e.g. >> motion parameter, brain volume, age,...). >> >> I arranged it as described below (or have a look at the attachment): >> >> EV1 EV2 EV3 EV4.... >> 1 0 2 1 >> 1 0 -1 -3 >> 1 0 -3 5 >> 1 0 4 2 >> 1 0 5 4 >> 0 1 -2 -1 >> 0 1 3 -2 >> 0 1 -4 -5 >> 0 1 1 -4 >> 0 1 -5 3 >> >> EV1 codes for Group1 and EV2 for Group 2, while EV3 and EV4 are some >> covariates (demeaned). After doing melodic, I ran: >> >> dual_regression /somewhere/melodic_IC 1 -1 0 /somewhere/output `cat >> /somewhere/something.txt` >> >> Subsequently, I ran randomise with the components of potential interest. >> >> Now, I used the attached design.mat and design.con files and got my >> tfce_corrp_tstat images, indicating the group differences. Till now, >> everything was fine. >> >> However, when adjusting my .mat and .con files (e.g. deleting age from >> the covariate list..thus no longer correcting for age effects), I noticed, >> that the resulting tfce_corrp_tstat outputs were absolutely identical! >> >> I repeated this for several combinations of covariates, and even without >> putting any covariate into the analysis, but still all the outputs were >> identical. I have no clue on what is going wrong, but there seems to be a >> problem in how I am processing my data. >> >> >> PS: I also tried to define 2 groups in a design.grp file withe 2 columns >> for each EV, but this also ended up with the identical output images. >> > > > > -- > *Himanshu Joshi* > M.Tech. Cognitive & Neuroscience. > Senior Research Fellow > Department of Psychiatry > NIMHANS, Bangalore > > http://mbial.weebly.com/himanshu-joshi.html > > > > ------------------------------------------------------------------------------------------------ > > ------------------------------------------------------------------------------------------------ > Forschungszentrum Juelich GmbH > 52425 Juelich > Sitz der Gesellschaft: Juelich > Eingetragen im Handelsregister des Amtsgerichts Dueren Nr. HR B 3498 > Vorsitzender des Aufsichtsrats: MinDir Dr. Karl Eugen Huthmacher > Geschaeftsfuehrung: Prof. Dr. Achim Bachem (Vorsitzender), > Karsten Beneke (stellv. Vorsitzender), Prof. Dr.-Ing. Harald Bolt, > Prof. Dr. Sebastian M. Schmidt > > ------------------------------------------------------------------------------------------------ > > ------------------------------------------------------------------------------------------------ > -- Cornelius Werner cornelius.werner<at>gmail.com