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 2131 - 2160 of 2632
3dttest++, 3dRegAna, and 3dMVM all should work for the situation, but the former two would be much faster. With 3dttest++ you need to put the covariate(s) into a text file and use option -covariates.
by
gang
-
AFNI Message Board
> If I use 3dttest++ I will have three coeff but I'm not sure I will get what I want.
Claudio, could you explain a little bit more what you mean by the "three coeff"? You only have one correlation result from each subject, correct?
by
gang
-
AFNI Message Board
Try this:
3dFDR -addFDR -FDRmask roi_subject+orig stats.subject+tlrc
The command would computes the FDR curve for all the statistic sub-brick including #133 you're interested, and store the information in the header of the file. On the AFNI viewer you would be able to see the FDR q-value right under the threshold slider bar.
by
gang
-
AFNI Message Board
I just modified 3dMEMA so that 3dMEMA could take 1D input files as follows (notice that the output file name has to have an affix of 1D):
3dMEMA -prefix bothLOC.hrf2.Acc.1D -jobs 1 -set A MRI_Subj_1 3dMEMA_S1_coef.1D 3dMEMA_S1_T.1D \
MRI_Subj_2 3dMEMA_S2_coef.1D 3dMEMA_S2_T.1D \
MRI_Subj_3 3dMEMA_S3_coef.1D 3dMEMA_S3_T.1D \
MRI_Subj_4 3dMEMA_S4_coef.1D 3dMEMA_S4_T.1D \
MRI_Subj_5 3dMEM
by
gang
-
AFNI Message Board
Hi Ying,
I manually ran the analysis and got the following result for you:
Estimate SE T-stat DF P-val
-4.5801 0.8549 -5.3573 9 0.000458
by
gang
-
AFNI Message Board
No, I was asking you to provide your *original* script with numbers as shown below, not the one with the 1D files.
3dMEMA -prefix result -cio \
-set A \
S1 4.2 3.3 \
S2 3.5 2.2 \
...
by
gang
-
AFNI Message Board
Hi Ying,
Does each subject have only one effect estimate plus its t-stat at the ROI?
Actually I'm not so sure if 3dMEMA works with 1D files or not. Could you paste the full version of your original script?
3dMEMA -prefix result -cio \
-set A \
S1 4.2 3.3 \
S2 3.5 2.2 \
...
by
gang
-
AFNI Message Board
> response regressors, are the same across different dimensions, I get a collinearity warning during
> the execution of 3dDeconvolve. Would this result in instability of coefficients of any of the regressors?
It's possible but hard to tell in abstract sense. You can take a close look at the results and see if they are reasonable. If they are, then you're fine.
> If I am
by
gang
-
AFNI Message Board
> For the contrasts from GLT I do not have the sub-bricks: hit13LibvsBase#1_Coef
> and hit13LibvsBase#1_Stat". (I only have those in the first part of the output that you indicated correctly).
Since I can't see your file GLTs4/hit13liberal.glt, my guess is that you only requested for the contrast of the average, not the marginal, effect. You would have to add a separate tes
by
gang
-
AFNI Message Board
So what you mentioned before (shown below) is *not* exactly what is in the output?
hit#0_Coef
hit#0_Stat
hit#1_Coef
hit#1_Stat
hitvsBase#0_Coef
hitvsBase#0_Stat
If I have to guess, this is how to interpret the result. Use condition 'Liberal13hits' as an example, you should see something like the following"
Liberal13hits#0_Coef
Liberal13hits#0_Stat
Liberal13hi
by
gang
-
AFNI Message Board
> I am not sure if this is a power issue
Several possible sources that could contribute to the detection discrepancy, and it's difficult to pin down precisely the exact the reason. For example,
1) not enough number of subjects;
2) suboptimal spatial alignment across modality/subjects;
3) some problematic PPI steps (e.g., deconvolve via 3dTfitter is imperfect).
Even if a clust
by
gang
-
AFNI Message Board
> It is the file that is output when making my interaction regressor in step 4 with the output of 3dTfitter
> and my coded stim file which has 1,0,and -1's. That output file has -0's. Is that just because some of
> the values in the output from 3dTfitter are so small?
That's most likely the reason.
> I passed it through the rest of the pipeline up to the group a
by
gang
-
AFNI Message Board
> Would you suggest dropping the last trial (letting it be a 0) or do I just simply need to
> extend the run_len to 316.01 (this stim time would only occur in the last run so I don't
> think it would affect any other timing) and delete the last line after the fact.
You can simply drop that last trial because there is no information about the trial in the signal.
> just t
by
gang
-
AFNI Message Board
If exposure is a within-subject factor, you have a 3 x 2 mixed factorial ANOVA: one between-subjects factor genotype (3 levels) and one within-subject factor with two levels. You can use 3dMVM for it.
by
gang
-
AFNI Message Board
I assume that you want to analyze resting state and emotional processing separately at group level.
> exposure also has two levels. Within each group there are not between subject variablity
So exposure is a within-subject factor? In other words, you have two z-scores and two betas from each subject, right?
by
gang
-
AFNI Message Board
Noah,
Could you clarify the number and types of explanatory variables you're considering?
1) Genotype is a between-subjects factors with three levels. Do the three genotypes have equal number of subjects?
2) Is imaging result a within-subject factor with two levels: resting state functional connectivity and emotional processing?
3) Is exposure status another factor with two level
by
gang
-
AFNI Message Board
Matt, the bug gave incorrect denominator degrees of freedom for the F-statistics when only between-subjects factors are present in the model. And this only occurred in the most recent version I released three weeks. It does not look like your case. Plus the bug has nothing to do t-test directionality.
It's still not clear to me exactly what issue you have. I'll send the updated versi
by
gang
-
AFNI Message Board
> Are these variants of double gamma?
No. You can find the details about various options for response functions in the help
3dDeconvolve -help | less
by
gang
-
AFNI Message Board
Hi Matt,
What version of AFNI are you using? I just realized that there is a bug in the most recent version of 3dMVM that may impact on the F-stat if all factors are between-subjects. But it should not have any effect on t-stat. The two different results about GLT were run with different versions of 3dMVM?
by
gang
-
AFNI Message Board
> I was wondering whether anyone can advise on what model to use during deconvolution if i need
> to model an event of 1sec duration with an undershoot. I am hesitating between SPMG1 and SPMG3...
As hemodynamic response may vary across trials, tasks, brain regions and subjects, there is no way to know in advance which model fits better than others. You may try out SPMG1, SPMG2, SPMG3,
by
gang
-
AFNI Message Board
> When I am viewing my data, I should select the same sub-brik for the overlay and threshold, correct?
Yes, that would make sense because, unlike a t-test, there is no simple effect estimate (response magnitude) associated an F-test.
by
gang
-
AFNI Message Board
> I was wondering if it would be more useful to apply multiple comparisons based on a certain Z-score value
Each z-score is directly associated with a specific p-value. In other words, if you know the z-score, you can find out the corresponding p-value. Using the z-score from the quoted FSL example, you can obtain the p-value for the Z-score of 2.3 with the following AFNI command:
cdf -t
by
gang
-
AFNI Message Board
You may have to change to a different fitting approximation. For example, try the following and see if it works:
3dTfitter -RHS ${subj}_clust1_anti_avgseries_seed_upsampled.1D -FALTUNG ${subj}_anti_GammaHR.1D ${subj}_anti_Seed_Neur 012 -1 -l2sqrtlasso 2
by
gang
-
AFNI Message Board
> I am curious if there is a way to correct for multiple comparisons using Guassian
> Random field theory methods instead.
Random field theory is currently not implemented in AFNI, so you'd have to try another package.
> Cluster-size methods are killing my results in smaller structures
First of all, random field theory is also a cluster-based approach for multiple testin
by
gang
-
AFNI Message Board
> How do I run a t-test with the values from the coefficient sub-brik for any particular contrast.
> I know SPSS (or AFNI) will report the average, and SPSS will report the stdev and SE.
> Would I then divided each voxel score by the SE to get a revised t-stat for the ROI?
Are you trying to obtain the average beta value within the ROI from each subject, and then run a t-test at gro
by
gang
-
AFNI Message Board
Is your factor A a between-subjects factor like groups ASD andTD? If so, you should use 3dANOVA3 -type 5 ... -clevels 4, and change the coding for the second group as follows:
-dset 2 1 1 'td01.LOC.4runs.GLM+tlrc[4]' \
-dset 2 2 1 'td01.LOC.4runs.GLM+tlrc[37]' \
-dset 2 1 2 'td02.LOC.4runs.GLM+tlrc[4]' \
-dset 2 2 2 'td02.LOC.4runs.GLM+tlrc[37]' \
by
gang
-
AFNI Message Board
Sure you can take the response amplitude at the ROI level from each subject, and run t-test with R/Excel/Matlab/SPSS.
> would I then be able to undump these values into the AFNI GUI.
I'm lost here. The t-stat at the ROI is just one number: Why would you want to undump the t-test at the ROI into the AFNI viewer?
by
gang
-
AFNI Message Board
Elisa,
I would like to have some clarifications from you before I could offer suggestions.
> I need a way to control for RTs in the ANOVA, that is, remove that part of variability
> that is due to differences in RTs across conditions. Now, having the ABI map that
> tells how activity in each region is modulated by RTs, how can I use it to remove,
> for each voxel, the modulat
by
gang
-
AFNI Message Board
> What are the values for the second (ABI) regressor? The value of the correlation between
> brain activity and RTs (positive in red and negative in blue)?
Yes, the second regressor reveals the response amplitude *change* when RT increase in one unit.
> I need to do a group analysis where I examine the interaction of criterion and item type
> *after removing* the effect of R
by
gang
-
AFNI Message Board