Hi, Gang.

Thanks again for the quick reply.

I've re-run my test analysis using FSL Level 1 (individual subject) COPEs as input to 3dLME. One set of COPEs is the output of a emot-fixation contrast. The other set is the output of a shape-fixation contrast.

Unfortunately, I don't think I understand the sub-brick ranges in the 3dLME output.

Here is the model submitted to 3dLME:

-model "cond*group" \
-ranEff '~1' \
-SS_type 3 \
-num_glt 3 \
-gltLabel 1 'emot-shape' -gltCode 1 'cond : 1*emot -1*shape' \
-gltLabel 2 'emot' -gltCode 2 'cond : 1*emot ' \
-gltLabel 3 'shape' -gltCode 3 'cond : 1*shape' \

And here is the output of 3dinfo from the new analysis:

Number of values stored at each pixel = 10
-- At sub-brick #0 '(Intercept) F' datum type is short: 0 to 32767 [internal]
[* 0.00305185] 0 to 100 [scaled]
statcode = fift; statpar = 1 18
-- At sub-brick #1 'cond F' datum type is short: 0 to 32767 [internal]
[* 0.00305185] 0 to 100 [scaled]
statcode = fift; statpar = 1 18
-- At sub-brick #2 'group F' datum type is short: 0 to 32767 [internal]
[* 0.000889488] 0 to 29.1458 [scaled]
statcode = fift; statpar = 1 18
-- At sub-brick #3 'cond:group F' datum type is short: 0 to 32767 [internal]
[* 0.000956772] 0 to 31.3505 [scaled]
statcode = fift; statpar = 1 18
-- At sub-brick #4 'emot-shape' datum type is short: -23879 to 32767 [internal]
[* 0.00305185] -72.8751 to 100 [scaled]
-- At sub-brick #5 'emot-shape Z' datum type is short: -18880 to 32767 [internal]
[* 0.000616474] -11.639 to 20.2 [scaled]
statcode = fizt
-- At sub-brick #6 'emot' datum type is short: -32767 to 32767 [internal]
[* 0.00305185] -100 to 100 [scaled]
-- At sub-brick #7 'emot Z' datum type is short: -23784 to 32767 [internal]
[* 0.00059687] -14.1959 to 19.5576 [scaled]
statcode = fizt
-- At sub-brick #8 'shape' datum type is short: -17084 to 32767 [internal]
[* 0.00290745] -49.6709 to 95.2685 [scaled]
-- At sub-brick #9 'shape Z' datum type is short: -22558 to 32767 [internal]
[* 0.000254238] -5.7351 to 8.33061 [scaled]
statcode = fizt

Several of the scaled ranges seem like they are at floor and/or ceiling. For example, sub-brick #1 'cond F' ranges from 0 to 100, and sub-brick #6 'emot' ranges from -100 to 100.

Do you have any insight into what might be going on?

In case it's helpful, the AFNI and R versions I'm using are:
Precompiled binary linux_xorg7_64: Jan 1 2016 (Version AFNI_16.0.00)
R version 3.2.1 (2015-06-18) -- "World-Famous Astronaut"

Also, the average range of values in the input COPEs is around -300 to 550.

Thanks in advance,
Ruskin