Dear SPM Users,
I am trying to implement an FIR analysis on my event-related data and was puzzled by the resulting design matrix output.
I performed the first-level analysis using the FIR basis function. TR = 2s, window length = 18s, Order = 9, no other parameters were changed. The resulting design matrix looks like FIR_X_autocorrelation (attached and shown below). I am puzzled why the design matrix isn't composed of simple diagonals for each trial, but seems to have some temporal derivative incorporated into the later delta functions.
I therefore tried the same thing again but without AR(1) to correct for serial correlations, and the resulting design matrix looks like FIR_X_noautocorrelation (attached and shown below). This looks much better, obvious clean diagonal matrix, but still, there seems to be negative modeled values in the later orders for each trial. Does anyone know what these "extra" values are? Am I missing some assumptions in the way SPM creates an FIR design matrix?
Any help would be greatly appreciated.
Thanks,
Josh Goh
