> I'm doing 2 t-tests first, and the 3rd glt is based on the all_cond.mat, which looks like
> 0 0 0 1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1 -1 -1 0 0 0 0 0 0 0 0 0
> 0 0 0 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 -1 -1 -1 -1 -1 -1 -1 -1 -1
>
> is this the right way to do a single-subject ANOVA test?

First of all you can encode the above matrix in the following symbolic fashion

-gltsym 'SYM: +Low[0..8] -Medium[0..8] \ +Low[0..8] -High[0..8]' -glt_label 3 'three_levelsAUC' \

If you ONLY care about the difference of the AREA under the (hemodynamic response) curve among the three conditions, yes, such an F-test would be what you want. However, there are two limitations associated with this approach: (1) two HRF's might have totally different shapes but with the same area under the curve; (2) it's even more problematic if you have some negative beta's among those coefficients.

An alternative approach would be more reliable and enable you to detect the subtle difference of the hemodynamic response SHAPE across those 3 conditions:

-gltsym 'SYM: +Low[[0..8]] -Medium[[0..8]] \ +Low[[0..8]] -High[[0..8]]' -glt_label 3 'three_levels' \

which is equivalent to the following glt matrix

0 0 0 1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
...
0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0
...
0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1

Be realized that the same limitations and alternative approach equally apply to your first two glt tests as well.

HTH,
Gang