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 2371 - 2400 of 2632
The terms "fixed-effects" vs "random-effects" are a twisted mess in the FMRI field that started from the early days.
A random-effects model requires multiple measurements from each subject so that the deviations of each subject from the group average can be modeled as a cross-subjects variable. In this sense the traditional ANOVA with at least one within-subject (or repeat
by
gang
-
AFNI Message Board
Hi Paul,
Nice to see you showing up here!
> in looking at Gang's page, I notice that you're declaring the previous method outdated.
That's why I'm not Obi-Wan Kenobi.
> I wonder if you'd mind directing me and others to the state of the art?
If you like Rick's the GUI pipeline on uber_subject.py, there is an option that toggles between 'task&
by
gang
-
AFNI Message Board
> Can anyone please tell me where to find out what the contrast options mean, such as tau^2 and QE?
tau^2 is the cross-subjects variance and measures the heterogeneity of the group, and QE is the chi-square test that indicates whether the cross-subjects variance is significant at a brain region.
If you run the analysis with the assumption of different variance across the groups (with op
by
gang
-
AFNI Message Board
I still think that you should use 3dMVM (not 3dLME) with something like
-model 'condition1*condition2+q1score+q2score+RT+Age' \
by
gang
-
AFNI Message Board
For your 2 x 3 ANOVA, the significant interaction between factors A and B basically means at least one of the following three composite comparisons is significant:
A1(B1-B2) - A2(B1-B2)
A1(B1-B3) - A2(B1-B3)
A1(B2-B3) - A2(B2-B3)
3dLME can give you those 6 within-group comparisons: A1(B1-B2), A2(B1-B2), ..., but unfortunately it currently does not allow one to directly obtain the above th
by
gang
-
AFNI Message Board
3dLME can handle missing data at some factor levels, but it can't deal with missing covariate values. You have two options: either don't use those subjects with missing covariate values, or don't include the covariate. Regardless, it seems that you should use 3dMVM, not 3dLME.
by
gang
-
AFNI Message Board
Now I see what you mean. Do you mind uploading all the input files? Thanks!
by
gang
-
AFNI Message Board
That sounds like an erratic phenomenon!
If you have a Dropbox account, could you send me the results from both 3dMVM and 3dttest++?
by
gang
-
AFNI Message Board
> When I created the contrast as you saw in my script, only regions which
> had shown a single directional change (ie, yellow) in the ttest output, and
> thus a subset of the clusters identified in the "experiment" main effect
> were present in the contrast output.
By "yellow" clusters, do you mean that only the positive clusters show up? Do you get exactl
by
gang
-
AFNI Message Board
> Yet, when I looked, the post hoc (gltLabel 1) only reported the upregulated regions.
What do you mean by 'upregulated regions'? Could you elaborate a little more about this?
by
gang
-
AFNI Message Board
> I was wondering whether there is a way to force zero at the first TR but not the last?
Easy! Suppose that your original basis function is specified as CSPLINzero(0, END, n), and the grid length of your basis function is L (e.g., 1 TR). Now you just change it to CSPLINzero(0, END+L, n+1).
by
gang
-
AFNI Message Board
Two potential issues that may have given the grief:
1) If variable 'colony' is quantitative, it should be declared with option '-qVars'.
2) if variables 'experiment' and 'time' are within-subject (repeated-measures) factors, they need to be specified with option '-wsVars'
by
gang
-
AFNI Message Board
> what if i'm comparing within-group (paired): conditions (A-B) vs. (C-D)?
Hi x, do you mean A, B, C, and D are four within-subject conditions? If so, you can obtain A - B - C + D from each subject, and then run 3dMEMA. You may want to run 3dMEMA on A - B and C - D separately to sort out the directionality.
by
gang
-
AFNI Message Board
Eliana,
You can try FDR correction by running the following
3drefit -addFDR -FDRmask InputFile
Regarding family-wise error (FWE) correction, it's a little dicey. On one hand, FWE correction assumes Gaussian distribution of the data, so it may contradict with the underpinnings of the Monte Carlo simulations with 3dClustSim. However, the output from the nonparametric programs is conve
by
gang
-
AFNI Message Board
Both versions of AFNI look too old. Follow Peter's suggestion of updating your AFNI, and see if the issues would be resolved.
by
gang
-
AFNI Message Board
Hi Rengin, the "-conditions" option has been disabled in 3dMEMA because a contrast between two conditions is better modeled by taking the contrast directly from individual subject level.
by
gang
-
AFNI Message Board
> for some reason the posthocs all look exactly the same so it seems something is wrong....
I can't see anything obviously wrong with the script. Could you contact Brian Pittman (pittmanb@mail.nih.gov) about the procedure to upload your data?
> I've been running another version of ths analysis with two covariates (instead of one)
> and it looks like it will take about
by
gang
-
AFNI Message Board
The issue of unpooled vs. pooled in 3dttest++ is about the assumption of whether the two groups have roughly the same variability. It's hard to tell beforehand which is more appropriate. If this is of concern to you, try out both options, and see if significance difference exists.
Alternatively you can use 3dMEMA with option -unequal_variance to explore the difference in within-group vari
by
gang
-
AFNI Message Board
Such a scenario warrants a close look why it occurred: head motion? subject anticipation? coding mistake in stimulus timing? Just throw in an option. If you don't believe that the brain should respond at the stimulus onset time point, simply use TENTzero (or CSPLINzero) instead of TENT (or CSPLIN). However, if you do see some large amount of signal at the onset time, consider adding a regres
by
gang
-
AFNI Message Board
I've been asked to pass this along.
=========
Call for papers: Frontiers in Neuroinformatics special topic “Recent advances and the future generation of neuroinformatics infrastructure”
Deadline for Abstract Submission: 31 Jul 2013
Deadline for Article Submission: 31 Oct 2013
The huge volume of multi-modal neuroimaging data across different neuroscience communities has posed a d
by
gang
-
AFNI Message Board
Still hard to tell what triggered the failure. Do you have a Dropbox account? If yes, could you send me all the input files plus the 3dMVM script?
by
gang
-
AFNI Message Board
Could you post the text between the following two lines on screen when you run 3dMVM?
***** Summary information of data structure *****
***** End of data structure information *****
by
gang
-
AFNI Message Board
Do you happen to have a Dropbox account? If so, could you send me the 3dLME output file plus the mask file?
by
gang
-
AFNI Message Board
Andrew, what version do you have for 3dLME? You can check it by running the following on terminal:
3dLME | head -10
by
gang
-
AFNI Message Board
> Do I remove residual in the 3dcalc step to get the seed time series?
No, you do need to keep the residuals in the final result of the seed time series because the residuals presumably contain something that is of interest for the correlation analysis.
by
gang
-
AFNI Message Board
How many time points does the variable 'time' have? The number of subjects in each group should be greater than the number of time points. If not, the system is not solvable, and you would have to use 3dLME.
by
gang
-
AFNI Message Board
> Am I correct in understanding that the following script models Group*Condition*Time?
Right.
> However, I keep getting this error. Thanks for the help!
You forgot to change the program name from 3dLME to 3dMVM in your script. Also it seems you need to make the following modifications since Age is a between-, *not* within-, subject factor:
-model 'Group*Age' \
-wsVa
by
gang
-
AFNI Message Board
The upload system on the AFNI server is currently broken. Is dropbox an option for you? If yes, send me a private message with your email address.
by
gang
-
AFNI Message Board
Matt, it looks like I'll have to ask you to upload all the input data plus the 3dMVM script to our server so that I can take a closet look at the situation.
by
gang
-
AFNI Message Board
It looks like the data can be better analyzed with 3dMVM. See Example 3 in the help:
3dMVM | less
by
gang
-
AFNI Message Board