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 1981 - 2010 of 2632
Barbara,
Thanks for the clarifications. If I understand your situation accurately, every subject has those five pipelines. If so, you should use 3dFriedman instead of 3dKruskalWallis.
I'm still not quite sure about the issue of zero values. Do you mean that you want to treat them as missing (instead of real) values?
by
gang
-
AFNI Message Board
Barbara, 3dKruskalWallis is meant to be used for comparing multiple groups of subjects. What are those five pipelines and what are those eight sub-bricks?
by
gang
-
AFNI Message Board
John,
Two points here. 1) If you run simple correlation analysis with a seed region, not everybody is keen on the approach, and you may have trouble with some reviewers. 2) It's not clear to me what your question is about regarding the multiple regressors for the correlation analysis. More importantly, I can't tell what question or issue you're trying to address with the analysi
by
gang
-
AFNI Message Board
> I'm thinking the no-go trials would have the amplitude modulation. Is this correct?
Unless this has already been investigated, nobody really knows whether or how one trial type has impact on the other trial type depending on the number of trials in between. You can try using the number of go-trials between two consecutive no-go-trials as a modulator and see if such a correlation exis
by
gang
-
AFNI Message Board
> I am curious about an interaction between base pair length in one gene and the base pair length in another (or similar idea).
By interaction do you mean that you want to see if the two genes have different correlation between connectivity and base pair length?
Assuming that each subject has different base pair length between the two genes, you need to use linear mixed-effects modeling
by
gang
-
AFNI Message Board
> is there a way to display it as negative linearity?
Display the negative results in what sense? At the ROI level? If so, you can extract the ROI connectivity measure from each subject, and plot them against the base pairs.
by
gang
-
AFNI Message Board
> When I see an area has increased connectivity, what is the correct way to interpret these results?
> Higher base pairs are associated with increased connectivity? Same thing with areas with
> reduced connectivity?
The effect magnitude (beta value in the output) means the amount of connectivity change (positive or negative) when base pair increases by 1.
> My guess is that
by
gang
-
AFNI Message Board
Matt,
You can use 3dMEMA in a piecemeal fashion: take each contrast (e.g., 1- vs. 0-back, 2- vs. 0-back, and 2- vs. 1-back) as input. However, this would not be able to give you an omnibus test for the overall interaction between the two factors if that's is part of what you're looking for.
by
gang
-
AFNI Message Board
The difference in those post hoc tests is due to the different testing strategy and assumptions. Specifically GroupAna makes a stronger assumption and uses a pooled variance.
by
gang
-
AFNI Message Board
Do the omnibus F-stats match between the two programs?
The differences in post hoc tests are most likely due to the different modeling strategy. For example, for the following test
-gltLabel 1 PT-HC_v1_C -gltCode 1 'group : 1*PT -1*HC visit : 1*v1 cue : 1*circ' \
does 3dMVM render the same t-value as GroupAna? Furthermore, check out their degrees of freedom for both t-stats: I
by
gang
-
AFNI Message Board
> is it more reasonable to compare the amplitude? And would the gltsym be like "A[0] -B[0]" ?
That's what most people would do. In addition, you could add an F-test
-gltsym 'SYM: A[0] -B[0] \ A[1] -B[1] '
which shows at least one of the two components differs between the two conditions.
P.S.: corrected per Rick's response below!
by
gang
-
AFNI Message Board
> Let's say 21 subjects in both the patients and control groups, respectively. However, for the
> patient group, one subject has data missing data for the neutral condition, and two subjects
> have data missing for the positive condition: How can one do pairwise contrasts between pos
> and neutral conditions across the group, when you have missing values in each condition.
by
gang
-
AFNI Message Board
Which part of 3dMVM output are you comparing with which part of 3dttest++?
Remove the option -unpooled in 3dttest++, and do the comparison again.
by
gang
-
AFNI Message Board
Rock, actually I was wrong in both of my previous responses!
I meant to say that the stimulus duration does matter when creating the interaction regressor in step 5 (not 4) where the re-convolved time series is filtered through 0s and 1s.
by
gang
-
AFNI Message Board
Actually I was wrong (kindly pointed out by someone through personal email). You do need to modify the script (step 4 on my website) when you convolve the derived signal at the neuronal level with the stimulus information (which contains stimulus durations). Sorry about my mistake.
by
gang
-
AFNI Message Board
No, the trial duration does not play any role in terms of generating the PPI regressor.
by
gang
-
AFNI Message Board
> I want to know if I need to do the fisher-z translate for the correlation coefficient.
The transformation is usually recommended to meet the normality assumption of the Student's t-test. You can do so with the following command:
3dcalc -a Corr_subj01R+tlrc -expr 'log((1+a)/(1-a))/2' -prefix Corr_subj01_Z
by
gang
-
AFNI Message Board
Ambrose, the message board only allows for attaching image files. You may have to paste the script here.
Incorporating a covariate usually would not change the significance much unless there is some strange distribution about the variable. How are the covariate values distributed? Well centered, or skewed?
by
gang
-
AFNI Message Board
It's difficult to gauge the situation. However, with task-based data, PPI is what is typically done:
by
gang
-
AFNI Message Board
The warning seems to just alert you that all the input files have the same name. Since you know that they are different files, you should be fine.
by
gang
-
AFNI Message Board
Is it resting state data? One or two groups? It's difficult to assess the situation without access to the data...
by
gang
-
AFNI Message Board
> we do not see statistically significant results at the seed region in the group
> output for either a) the interaction coefficient, or b) the seed time series
> coefficient. Does that make sense?
Just ignore the situation at the seed region because it's not really interpretable.
by
gang
-
AFNI Message Board
For the analysis with covariate, did you center the covariate properly? And how does the result compare to the one without the covariate?
by
gang
-
AFNI Message Board
> we noticed in our group level t-test on the Seed_ts parameter that we are not seeing
> high correlations (or a high beta value) in the tlrc BA25 region (it's actually surprisingly low).
Do you mean that you don't see statistically significant results at the seed region in the group output? Other than this, are you generally happy about the whole analysis? If so, don't
by
gang
-
AFNI Message Board
Megan,
The results from option -wsMVT should be used as complementary ones to the default F-stats. In other words, they are useful only if you see any statistically significant regions that fail to show up in the conventional F-stats.
> in the view the autorange for each of these F sub bricks is 0 to %
This is strange. Do you mind provide the output of the following command?
3dinfo
by
gang
-
AFNI Message Board
Maybe you can start off doing some exploratory analysis with the InstaCorr stuff before settling down a specific analysis:
by
gang
-
AFNI Message Board
Paul, are you running on a 1x1x1 resolution? You could resample the data during the step you align the individual subject data to standard space at a resolution close to your EPI data (e.g., 3x3x3), and that would save a lot of runtime.
by
gang
-
AFNI Message Board
Are those spikes at the censored time points? If so, it should not matter either way because you should censor those time points again later in your 3dDeconvolve script.
by
gang
-
AFNI Message Board
If the first 3 TRs have already been removed from the file, use all_runs.subject+tlrc'[0..238]' ; otherwise go with all_runs.subject+tlrc'[3..241]'.
by
gang
-
AFNI Message Board
> For the option you suggested, would I use the TR timing or seconds?
Use the TR grids. You don't know the number of TRs for each run?
by
gang
-
AFNI Message Board