Show all posts by user
Dear AFNI users-
We are very pleased to announce that the new AFNI Message Board framework is up! Please join us at:
https://discuss.afni.nimh.nih.gov
Existing user accounts have been migrated, so returning users can login by requesting a password reset. New users can create accounts, as well, through a standard account creation process. Please note that these setup emails might initially go to spam folders (esp. for NIH users!), so please check those locations in the beginning.
The current Message Board discussion threads have been migrated to the new framework. The current Message Board will remain visible, but read-only, for a little while.
Sincerely,
AFNI HQ
History of AFNI updates
Results 2341 - 2370 of 2632
> In my case, would I "marry" the number of stimuli to the event time I have
> listed in my original timing files? So for example, if a memory array of set
> size 2 occured at 16 seconds. In my timing file I would put 16*2?
Yes, that's right. The marriage (or divorce) can be achieved with 1dMarry:
1dMarry | less
by
gang
-
AFNI Message Board
Is the number of stimuli associated with each trial event? If so, you may use option -stim_times_AM2 in 3dDeconvolve. See the document for more details:
3dDeconvolve -help | less
by
gang
-
AFNI Message Board
Hi Cindy,
1) Make sure your path is set properly:
2) update your AFNI at the terminal:
@update.afni.binaries -d
3) Use 3dClustSim. AlphaSim is too old and not as efficient as 3dClustSim.
by
gang
-
AFNI Message Board
> I hate matlab
We share the same perception.
> you have abandoned GroupAna
I no longer provide any support for it, but the package is hanging around.
> The version of 3dMVM that I use is 0.1.8 (Jan, 2013). It is a little bit old.
Yes, update your AFNI first and see if that fixes the problem:
@update.afni.binaries -d
by
gang
-
AFNI Message Board
The error message is very obscure. Could you provide the 3dMVM script?
by
gang
-
AFNI Message Board
> How can I threshold the F subricks without the T-statistic (like we usually get in 3dANOVA)?
Do you mean that you want to know the directionality of the covariate effect? If so, 3dLME should automatically output the effect plus the associated t-statistic.
> I have a few other covariates that are similar to age and in one case the groups
> definitely do differ. Would that be a pr
by
gang
-
AFNI Message Board
No, the file attachment didn't work. Is it in bmp, gif, or jpg format? And the file size has to be smaller than 614KB. You may also send it directly to my mailbox.
by
gang
-
AFNI Message Board
Hi Andy,
My personal opinion is that the assumption of homogeneity of regression for ANCOVA is really an old school stuff. Without knowing (or testing) whether the groups have the same covariate effect, a prior assumption seems a little foolhardy.
Having said that, 3dttest++ and 3dMEMA both have the capability to model the same or different covariate effect across groups. 3dMEMA has the opt
by
gang
-
AFNI Message Board
Hi Erin,
Are all of your potential covariates at subject level like age? In your current case you can use 3dttest++, 3dMEMA, 3dMVM, or 3dLME. In terms of usage, 3dMVM is probably a little easier than the other options including 3dLME.
For your 3dLME script, do the two groups have roughly the same average age? The script looks fine, and 3dLME (and 3dMVM) automatically outputs F-statistics fo
by
gang
-
AFNI Message Board
Hi Ting, I'm not so sure what caused the error, but option "-ranEFF" should be "-ranEff".
by
gang
-
AFNI Message Board
Considering such regression coefficients are enormously large, I suspect those are probably due to singularities (numerical artifacts). If those voxels are pretty isolated (not in clusters), I would not worry too much about them.
by
gang
-
AFNI Message Board
> I want to compute the sum of squares (or absolute sum) of a single volume (stored as one nifti file)
Use 3dcalc to square the data first, then 3dBrickStat -sum (with option -mask if you want to filter out the outside of the brain) should give you the result. In case you want to remove the mean before the squaring, use 3dBrickStat -mean ... -mask ...
by
gang
-
AFNI Message Board
Does it only happen with small values? If so, don't sweat it. For such small values, my guess is that the sign flipping is due to the truncation phenomenon when storing the data in short, for example.
by
gang
-
AFNI Message Board
I tried your data with 3dttest++, 3dMEMA, and 3dMannWhitney, and the three results are pretty much compatible with each other. There are two reasons for the lack of power in the group difference:
1) the number of subjects in the second group is too small;
2) The cross-subjects variability in the second group is way much bigger than the first one, and this is especially demonstrated from the 3
by
gang
-
AFNI Message Board
Yes, they should have the same sign, but could you show a pair of beta and t that have different sign?
by
gang
-
AFNI Message Board
This is puzzling to me! Could you send me the input files through Dropbox? Or drop me a message to my email and I'll give me the file upload instructions.
by
gang
-
AFNI Message Board
In your case with 3dttest++ for group comparison, it should give you three results in one analysis: one group comparison and two individual group results. So do you still see strong activation for one group in the same 3dttest++ output? Could you post your 3dttest++ script?
The pooled vs. unpooled issue usually should not make a big difference. How different in terms of the number of subjects
by
gang
-
AFNI Message Board
> I only need to model the stimulus using 'BLOCK(16,1)', am I correct?
Yes, that's right. Also if you haven't, consider using uber_subject.py or afni_proc.py to streamline your analysis steps.
by
gang
-
AFNI Message Board
> Is there a way to compare two groups to see what areas are significantly greater in one
> group in comparison to another?
3dttest++ and 3dMEMA are the two standard methods for group comparison.
> I used 3dttest++ but I am not getting very good results even though it seems like I should be.
Results not good in what sense? When you lower the threshold for the t-statistic to a
by
gang
-
AFNI Message Board
Qing,
Are you treating the resting periods as baseline condition? If so, you're not supposed to explicitly model them because the baseline condition is by default fitted with polynomials in 3dDeconvolve/3dREMLfit.
by
gang
-
AFNI Message Board
Lyn, could you post your 3dLME (especially the gltCode part) and 3dMVM scripts?
by
gang
-
AFNI Message Board
Eric,
Could you send me the following file?
reduced.concat.all.cleanEPI_eig
by
gang
-
AFNI Message Board
Have you checked your gmail account? I sent to you earlier the program plus one suggestion for your 3dLME script.
by
gang
-
AFNI Message Board
> I think there may be a bug in 3dLME when it attempts to split the -qVars line.
Yes, I just spotted the bug after I saw Isabelle's first post. And the fix is available in the next AFNI build. A temporary solution, as you've pointed out, is to code multiple quantitative variables with "-qVars 'var1+var2'" instead of "-qVars 'var1,var2'" (as
by
gang
-
AFNI Message Board
> there were several expected F values, but then also an odd t stat for only one of the covariates, as well as several “NA”.
Odd t-statistic in what sense? Several NAs for what?
For your 3dMVM script, what happened when you ran it with the following model?
-model "bcat1*bcat2+RT+Age" \
by
gang
-
AFNI Message Board
Eric, if you send me the files plus the 3dICA script through, for example, dropbox, I can take a loot at the situation.
by
gang
-
AFNI Message Board
> My plan is to regress out the task, and then calculate temporal correlations in the
> residual using 3dDeconvolve.
The regressing out step is not ideal, and can only remove the assumed response that is typically averaged across many trials. In other words, it won't touch the following: 1) the varying response across trials from the averaged curve; 2) the response that is differe
by
gang
-
AFNI Message Board
First time to see such an error message!
How big are your files? And what's the memory size of your system?
by
gang
-
AFNI Message Board
Hi Lyn,
Sorry I failed to associate the current context with what we chatted during the bootcamp.
Regarding the 3dLME script:
1) If 'bcat1' is 2-level categorical variable, use letters, not 1/0, to code it.
2) change the model line from
-model 'wnf1*wnf2*bcat1*bcon1+RT+Age'
to
-model 'wnf1*wnf2*bcat1+bcon1+RT+Age'
because you don't want a
by
gang
-
AFNI Message Board
Lyn, 3dMVM can run multiple regression without any problem. Alternatively 3dttest++ or 3dMEMA can do that as well. And 3dLME actually is not meant to be used for multiple regression.
by
gang
-
AFNI Message Board