Hi Nicola,
This is just a little confusing because the iresp options
have slightly different behavior than the betas, and it is
the betas that correspond to the gltsym indices.
The -iresp option shows a TR-locked time series (unless you
specify otherwise) for the duration of the CSPLIN. But the
betas (and regressors), which correspond to the stim class
indices in the GLTs, are as you specify with the parameters.
In this case I will assume your TR = 3s, since you say the
iresp output has 6 time points. But there are still only
4 regressors in the design matrix, because you used the
'zero' version of the basis function name. And since there
are only 4 regressors, the index of 4 is out of range, and
should give you an error message probably like:
** ERROR: selector index 4 is out of range 0..3
Though the command will still run (that is probably a
mistake, it should probably fail).
Yes, you should change 2..4 to 1..3 because the actual
basis function that you are using is CSPLIN(3,12,4). This
sort of confusion is why I suggest not using 'zero'
versions of these basis functions. They are just a
programming convenience to assume the end points
are zero, rather than estimating them.
Forget +1 and such. I suggest not using CSPLINzero,
but using CSPLIN the way your really want to. Do you
want CSPLIN(3,12,4)? If so, use that instead.
Does this make sense?
- rick