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 2101 - 2130 of 2632
> Do the extracted beta weights from 3dMVM generally include any contribution/influence from the covariates?
What beta weights are you referring to? Do you mean the effect estimate associate with -gltCode?
by
gang
-
AFNI Message Board
> is there a way to run 3dMEMA on multiple parameters and get an F-statistic?
No, 3dMEMA does not allow for multiple inputs from each subject. So the results from 3dMVM are not powerful enough?
by
gang
-
AFNI Message Board
Modeling the covariates most likely would not change the big picture. If you run the analysis with those two quantitative covariates, do the following:
1) perform within-group centering before you feed the covariate data into 3dMVM;
2) analyze the ROI data in R or SPSS with the same model as you do in 3dMVM.
by
gang
-
AFNI Message Board
> How does the official spm software handle this? maybe there are some hard
> constraints how much of the temporal/spatial derivative can be added.
The three basis functions coded in option SPMG3 of 3dDeconvolve are exactly the same as implemented in the SPM. Over-fitting is the caveat that can't be easily avoided with the approach of multiple basis functions since little constrain
by
gang
-
AFNI Message Board
> However, given that the units I am comparing are Fisher-transformed z-scores
> of correllations, I don't know how to make sense of the magnitude of these
> results. How can I interpret these units?
You can convert those z-scores back to correlation coefficients through the following
3dcalc -a Grp_Result+tlrc'[0]' -expr '(exp(2*a)-1)/(exp(2*a)+1)' -pre
by
gang
-
AFNI Message Board
> It works great now.
Thanks for the update!
> Unfortunately I seem to have too few subjects for multivariate analysis (I still dont
> quite get how degrees of freedom are obtained), eventually I would have 4 groups
> with ~6 subjects each.
The requirement for the sample size is
n > m + q
where n is total number subjects, m is the total number of level combinat
by
gang
-
AFNI Message Board
Most likely it's a memory issue. If it still happens when you run 3dMVM after quitting all other programs, you may have to find a computer with a bigger memory size.
by
gang
-
AFNI Message Board
Sahar,
It seems that your EPI data were collected from a task-related experiment. Your approach is just simple correlation analysis. FYI, the typical connectivity analysis for the situation is performed as the following:
by
gang
-
AFNI Message Board
> Are 6 subjects enough?
No, I need more than 7 subjects for one group and more than 8 for two groups. Thanks!
by
gang
-
AFNI Message Board
> I thoughts its because I have too few subjects, but even with 14 subjects the error remains...
With 14 subjects it should work. I'm puzzled by the error message. The ANOVA-style approach may be more efficient, but the null hypothesis would be uninterpretable. This is a perfect situation where 3dMVM is designed for. Do you mind uploading all the input files so that I could take a clo
by
gang
-
AFNI Message Board
〉as far as I know 3dMVM requires at least
〉one between-subject variable, so I included 'Group'.
No, this is not true. Change
-model Group
to
-model 1
by
gang
-
AFNI Message Board
Thanks for sharing the information! Do you know which part in R 3.0 that failed R_io.so?
by
gang
-
AFNI Message Board
> Since we didn't use -qVarCenters what is the default?
Read the information for option -qVarCenters in the help of 3dMVM:
3dMVM -help | less
> We think that we just had differences because of the nature of the subject groups.
> We think the covariates may be driving the effects, is that likely?
If that's the case, it may make more sense to perform within-group ce
by
gang
-
AFNI Message Board
A bug caused this problem. If you can't wait for the next release, send me an email and I'll give you the fixed version.
by
gang
-
AFNI Message Board
> 1) No, the 2 groups have significantly different values on these covariates (IQ, SES).
This is a very delicate and serious situation! I see that you didn't use option -qVarCenters in your script, which means you're comparing the groups at the overall average IQ and SES across the groups, which may or may not make sense. This may also explain the discrepancy between the 3dMVM ou
by
gang
-
AFNI Message Board
1) Does each covariate have roughly the same average value across groups? You may want to run two-sample t-test for this.
2) Could you provide your 3dMVM script?
by
gang
-
AFNI Message Board
For multiple regression, 3dRegAna may work, but it's not as user-friendly as 3dttest++ and 3dMVM in which you treat those quantitative variables as covariates.
by
gang
-
AFNI Message Board
If you're just comparing the two groups without any other covariates, you should not code the two groups as a covariate because the program automatically takes care of the situation.
by
gang
-
AFNI Message Board
Linda,
The ICA tool in AFNI is pretty basic. For group ICA, you'd be better off using FSL or GIFT (http://mialab.mrn.org/software/).
by
gang
-
AFNI Message Board
If you have varying durations across trials, the SPM basis functions are not really a good option. Instead you'd be better off using dmUBLOCK. See its usage in
3dDeconvolve -help
by
gang
-
AFNI Message Board
> can I use in 3dMEMA F-statistics associated with contrast instead of T-statistics?
No, F-statistics would not work because only t-statistics would provide proper information about the standard errors for the effect estimates. Could you explain why you'd like to use F-statistics?
by
gang
-
AFNI Message Board
> For specifying lag of one stimulus, if I use CSPLINzero(0,12,5) , does the first coefficient
> after the zero starting point (the one at delay 3 sec) counts as lag 0 or lag 1?
With CSPLINzero(0,12,5), the first coefficient is associated with the response at 3 s, corresponding to a lag of 1 in the old option. If your TR = 2 s, it'd be better to try CSPLINzero(0,12,7).
by
gang
-
AFNI Message Board
> as long as all possible linear combination of all rows can cover any
> contrast of any two conditions, and these rows are orthogonal to
> each other, then the f-statistics is fine?
You can set up the rows so that they are orthogonal with each other, but it's not really necessary for the tests you're interested.
> For example, if I have 4 levels, then a matrix as
by
gang
-
AFNI Message Board
The issue should have been avoided during the experiment design stage. You may try your luck by shortening the duration covered by the basis functions such as TENT(0,12,7), TENTzero(0,14,8), or TENTzero(0,12,7).
by
gang
-
AFNI Message Board
> Say I am contrasting two conditions, is the p-value based on the estimated beta of the
> two condition and the estimated standard deviation of the betas?
Right.
> (1) condition A is bigger than condition B AND condition A is bigger than condition C
Do two separate tests:
A vs B
A vs C
Then the intersection of A > B and A > C would be what you're looking f
by
gang
-
AFNI Message Board
> is there more or less connectivity between the amygdala and MFC in one group than the other
That's why I suggested in my previous response that you run an analysis for each group separately to tease apart the group difference. I don't know how your group analysis was set up with 3dLME, but I suspect that you should be able to obtain the within-group results with option -gltCode?
by
gang
-
AFNI Message Board
> Now I'd like to determine what is driving the differences, and this is where I am getting stuck.
> How do I figure out what is driving the differences in connectivity in the two groups?
Hi Colm,
You probably have already obtained the PPI effect for each group separately and that should help you interpret the group comparison. Other than that, I'm not so sure what exactly
by
gang
-
AFNI Message Board
> There is no file called just "3dMVM" in the bin directory.
That's weird because 3dMVM is just a text and should be automatically included in your binary directory. Try update your AFNI with the following command, and see if you can find it after the update:
@update.afni.binaries -d
by
gang
-
AFNI Message Board
3dMVM is supposed to run on the terminal via a script. Are you running 3dMVM on the Linux terminal or inside R? Also what is your AFNI version (2nd line from the output of the following command)?
afni -ver
by
gang
-
AFNI Message Board
Most likely there was not enough memory on your computer to run the analysis. Quit all other programs, and try it again. If it still fails, you may have to find a computer with a bigger memory.
by
gang
-
AFNI Message Board