Hi All,
Is the design matrix found in design.mat normalized in any way (other than being demeaned)?
I took my filtered_func_data and subtracted out each parameter estimate i.e.
e=filtered_func_data - mean(filtered_func_data) - PE1*reg1 - PE2*reg2 - ...
where reg1 is the first colum of the design matrix, reg2 is the second, etc
The result, e, should be the error in the model fit. The problem is that my sigma squareds are
about 3 orders of magnitude larger than the values in sigmasquareds.img. Or another way of
looking at it, each one of my modeled regressors i.e. PE1*reg1, is about 1-2 orders of magnitude
too small. After subtracting out the full fitted model, my error term looks almost identical to
filtered_func_data. So...
I suspect that I'm using the wrong design matrix values or perhaps there's a multiplicative factor
that I'm missing. Does anyone have any ideas of what I might be doing wrong?
thanks!
jack
|