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 2491 - 2520 of 2632
> abmean i j prefixname: is this the difference between level i of
> factor A and level j of factor B
No.
> is this command the difference between the sum of the means of
> level i of factor A and level j of factor B vs. the baseline condition?
"abmean i j" gives the group average and its t-statistic at level i of factor A and level j of factor B.
> I am t
by
gang
-
AFNI Message Board
Hi Mahen,
> I won't be doing any group analysis, and so far TENT and GAM are
> giving me very different BOLD response shapes. In terms of
> 3dDeconvolve results, the ROIs are fairly similar, with the TENT
> results seeming more robust. Considering that I am just using these
> results to 1) create masks, and 2) "clean up" my signal, are there
> any the
by
gang
-
AFNI Message Board
Hi Gaurav,
> would you recommend that I average the errts residuals
> from all subjects and feed that averaged residual to FWHMx
> program to find out the smoothness of noise at group level
> and feed that smoothness to 3dClustSim to figure out the
> cluster size threshold at group level?
The current practice is to estimate the FWHM using 3dFWHMx with the residuals fr
by
gang
-
AFNI Message Board
Hi Gaurav,
1. There are probably some inhomogeneities in terms of spatial correlation across the brain, and its impact is unknown if one simply smoothes the original EPI data with a uniform kernel size. However, if this is something of concern, you may consider adopting a uniform smoothing approach by using 3dBlurToFWHM, which would adaptively smooth the original data to achieve a uniform corr
by
gang
-
AFNI Message Board
Not so sure what's going on. Try changing the following option
-jobs 4 \
to
-jobs 1 \
and see how it goes.
by
gang
-
AFNI Message Board
> is there a way to measure R-Square for the whole brain instead of at a particular voxel?
Could you clarify this? The analysis with 3dRegAna is performed in voxelwise fashion in the sense each voxel is analyzed separately. What do you mean by R^2 for the whole brain?
by
gang
-
AFNI Message Board
> Is this R-Square for the whole model OR for a particular voxel where
> I have chose to click in AFNI overlay-underlay?
It's both, meaning that R^2 measures the amount of variability accounted for (relative to the total variability) by the model at each voxel.
by
gang
-
AFNI Message Board
> Does 3dMVM support within-subject covariates or is that still a 3dLME domain?
This is a very good question. Seriously!
3dMVM does not prevent one to run analysis with within-subject covariates. However, it is a delicate endeavor when dealing with within-subject covariates for the following two reasons:
1) How much does a within-subject covariate correlate with other variables, espe
by
gang
-
AFNI Message Board
> Would I still use PPI if I'm not interested in a specific task condition
> (i.e. I wouldn't be looking at connectivity of an ROI for a contrast between
> two conditions)?
PPI, in its typical usage, explores the interaction between a seed region and the rest of the brain under one condition or the contrast of two conditions.
> The simple correlation analysis ind
by
gang
-
AFNI Message Board
It seems that there are quite a few group analysis programs already in the AFNI suite: 3dttest++, 3dMEMA, 3dANOVAx, GroupAna, and 3dLME. Theoretically speaking, 3dLME could handle almost all of the complex situations of FMRI group analysis. However, there are a couple of thorny issues with the LME approach in practice even for some relatively simple analyses: flexible but difficult-to-standardize
by
gang
-
AFNI Message Board
> 1. When is it appropriate to use TENT/GAM/BLOCK basis functions
> (I'm primarily looking at the visual cortex in this study; 1s on, 15s fixation,
> with a response in the first second of those 15s)? I ask because I'm sure
> I've seen all 3 used in similar situations. I'm leaning toward using a TENT
> function here - probably TENT(0,16,9).
It all dep
by
gang
-
AFNI Message Board
As the understanding about perception and cognition deepens, the typical strategy of modeling the BOLD response with a presumed response curve with a fixed shape (same across brain regions, across all sorts of tasks or conditions, across subjects and groups) may become unable to reveal the nuances of brain responses. Instead accurately modeling the BOLD response could be necessary, and the adopti
by
gang
-
AFNI Message Board
Anita,
Instead of simple correlation analysis, the so-called PPI approach is generally considered for a task-related experiment in the field:
by
gang
-
AFNI Message Board
Gaurav,
Since you have a task-related experiment, you may consider the steps for the context-dependent correlation analysis here:
by
gang
-
AFNI Message Board
Hi Phil,
> My first question is, is this an accurate description of how to create this regressor?
Yes, it looks reasonable to me.
> My second question is, if this is an accurate description of the approach, why is it
> necessary to deconvolve the seed time course rather than convolving the condition
> code with the canonical HRF prior to multiplying by the seed? I suspec
by
gang
-
AFNI Message Board
3dLME should automatically generate F-statistics for main effects and interactions. So you don't need to make any request. Plus the following two lines are incorrect and unnecessary:
Group*Conditions
PD*Cond
Also you have a simple design, and can be easily analyzed with 3dMEMA or 3dttest++ without using 3dLME. See more discussion here:
by
gang
-
AFNI Message Board
> 1) How should I use homoskedasticity or heteroskedasticity,
> is heteroskedasticity the default type in 3dMEMA?
You found a bug in example 3 of the 3dMEMA help! The option -equal_variance (homoskedasticity) is the default, while option -unequal_variance corresponds to heteroskedasticity.
> 2) I read some response from Gang Chen "I tend to use
> -model_outliers and -un
by
gang
-
AFNI Message Board
> If you have more than 3 inputs (eg, 4) does the expression becomes:
>
> -expr 'step(a-4.2)+2*step(b-2.9)+4*step(c-3.1)+8*step4'
>
> or
>
> -expr 'step(a-4.2)+2*step(b-2.9)+4*step(c-3.1)+16*step4'?
Both are actually good enough to differentiate all the scenarios. However, the first one is better because the multipliers go by the rule of 2^n, n
by
gang
-
AFNI Message Board
> I already have ROIs defined, so I don't supposed I will begin with 1dGC and
> a single seed regions. Is this ok?
That's surely fine because 1dSVAR contains the functionality of 1dGC even if that's what you're looking for.
> I just wanted to confirm that it is possible to perform a second-level group
> analysis (using 3dMEMA, for example), as is alluded
by
gang
-
AFNI Message Board
I'd first try changing the model from 'TENT(0,14,8)' to 'TENTzero(0,14,8)' or 'TENTzero(0,12,7)'.
by
gang
-
AFNI Message Board
> I did find one hint online that it might be mean sum of squares,
> but I wasn't sure. Can you help?
You can pretty much ignore that sub-brick: it's the square root of mean sum of squares.
by
gang
-
AFNI Message Board
It seems that you're treating performance as a quantitative variable instead of a within-subject or repeated-measures factor. In other words, each subject has only one, not two, performance value. Your 3dRegAna script indicates that you're modeling the data with the following equation
Response = group_response + a1 * awareness + a2 * performance + a3 * awareness * performance + resid
by
gang
-
AFNI Message Board
> Awareness is continuous, and performance is discrete.
I assume this is group analysis. Is 'performance' a within-subject (or repeated-measures) variable? How many levels? Was 'awareness' measured at each performance level, or do you have only one 'awareness' value per subject?
by
gang
-
AFNI Message Board
> Our approach is to specify 7 stime files for the 6 task levels and the control
> condition. Then, we perform 6 GLM contrasts on each level vs. the scramble
> image control condition. The beta-coefficients of these contrasts are tested
> for differences.
This seems reasonable to me.
> A collaborator thinks there may be an alternate mechanism for specifying our
>
by
gang
-
AFNI Message Board
I'm still confused about what you're trying to achieve. If you want to comare the two groups of subjects, how about running a two-sample t-test with 3dtest++ or 3dMEMA, and then correct for multiple comparisons on the group difference? If you meant something else, please try to state exactly the comparison you're looking for.
by
gang
-
AFNI Message Board
> does the results differ between a) running trend analysis at
> individual level then putting the beta values into 3dttest++ for
> group analysis and b) putting all the subjects' data into
> 3dANOVA2 using -acontr option for trend analysis at a group level.
Yeah, I forgot about option b). The two options are expected to give you pretty close results. Either way should
by
gang
-
AFNI Message Board
> I am trying to compare two sets of data that have been thresholded
> by cluster size and significance (generated from alphasim).
Thresholded with each set's group analysis result?
> Is there a simple way to run comparisons (maybe 3dttest++?) between
> the two groups that includes cluster and significance thresholds?
It's not clear to me exactly what comparis
by
gang
-
AFNI Message Board
Hi Yi,
There are two approaches to handling your case:
1) At individual subject level, test the linear and quadratic trends among those four levels:
and then take the effect estimates (beta values) for the trends to group levels using 3dttest++ or 3dMEMA.
2) 3dLME.R can handle this directly at group level, but I'm still working on the interface.
by
gang
-
AFNI Message Board
SL Wrote:
-------------------------------------------------------
> I didn't get the second half of the following
> page. You meant it as the documentation on your
> website about 3dLME
> (http://afni.nimh.nih.gov/sscc/gangc/lme.html)?
Sorry, I forgot to include web page. Yes, that's what I was referring to.
> Also, in the case of having more contrasts - 2
&g
by
gang
-
AFNI Message Board