Hi Ruskin,

> I'm an FSL user, but would like to use 3dLME to address some complex statistical models.

Unless there are some missing data, you might just use 3dMVM in this case, which should give you pretty much the same results.

> 1) What type of data is assumed to be in the input files for 3dLME? Are they beta-weights, and, if so,
> would the AFNI version of that be analogous to a cope file in FSL?

Yes it takes effect estimates such as beta values (or COPE in FSL) as input.

> 2) When I request that the output be in nifti format (by appending .nii to the prefix name), the program
> produces a single nifti file with just one volume. This isn't a problem b/c I can use 3dAFNItoNIFTI to
> create separate nifti files from each sub-brick in the default brick output. I just thought I'd let you know.

That sounds like a bug with .nii as output. Thanks for reporting this. I may have to take a close look at the situation later.

> 3) In the output brick there are a number of sub-bricks, some of which appear to be statistical tests, but
> others of which are not labeled.

If you look closely, for each sub-brick there is a label, which is based on what the user specified.

In general, the output is structured like this in AFNI: Unlike F-statistics, each t-statistic is paired with the corresponding effect estimate. See more about this in my reply below. A side note related to this is that we emphasize the importance of reporting the effect estimates; that is, it is more informative to colorize the effect estimates, not the statistical values (which are only useful for the purpose of thresholding).

> I assume that sub-bricks 0-3 (statcode = fift) are the statistics from the omnibus f-test (so,
> sub-bricks 1-3 are the f-statistics associated with the two main effects and the interaction, in
> that order, correct?). However, sub-bricks 4 and 5 both refer to a single contrast. I think
> sub-brick 5 is a map of z-statistics (statcode = fizt), but I don't know what sub-brick 4 is.
> Can you enlighten me?

Use sub-brick 2 as an example:

-- At sub-brick #2 'group F' datum type is short: 0 to 32767 [internal]
[* 0.00100001] 0 to 32.7675 [scaled]
statcode = fift; statpar = 1 18

The label 'group F' is right after the sub-brick number '#2' in the first line. The whole label 'group F' comes from the variable name you provided: 'group'. The AFNI program simply attaches an 'F' to the variable name, indicating that this is an F-statistic volume.

You're correct that the statistical code 'fift' means F-statistic. Furthermore, the two numbers in the statistical parameters part (statpar = 1 18) show the numerator and denominator degrees of freedom for the F-statistic.

Here is another example:

-- At sub-brick #4 'emot-shape' datum type is short: -11961 to 32767 [internal]
[* 0.000223776] -2.67659 to 7.33248 [scaled]

The #4 sub-brick is labeled as emot-shape: a contrast between 'emot' and 'shape', within the range of the difference in [-2.67659, 7.33248]. I'm not so sure about the dimensionality here: it does not seem to be in percent signal change?

Gang