Dear AFNI friends,

I am trying to use the SPMG3 response model on some data in 3dDeconvolve or 3dREMLfit.

Usually I use the CSPLINzero response model, and then I look at a specific part of the area under the curve in a glt: eg. -gltsym ‘SYM: +MyStim[2…4]’

I’m wondering two things: What is the formula you use to create the -iresp from SPMG3?

And how would I set up a glt for certain times post stimulus onset for the SPMG3 estimates, like I do for CSPLINzero?

Thank you for your help, and all of your work on AFNI,

Sally

Hi Sally,

What is the formula you use to create the -iresp from SPMG3?

The option -iresp in 3dDeconvolve should automatically assemble the overall response curve by summing up the three regressors.

how would I set up a glt for certain times post stimulus onset for the SPMG3 estimates, like I do for CSPLINzero?

No easy way to do that at the individual level. If your goal is the group analysis, you can directly take the output from -iresp to group analysis.

May I suggest the following paper?

Chen, G., Saad, Z.S., Adleman, N.E., Leibenluft, E., Cox, R.W. (2015). Detecting the subtle shape differences in hemodynamic responses at the group level. Front. Neurosci., 26 October 2015. http://dx.doi.org/10.3389/fnins.2015.00375

Thanks, Gang,

The article was very helpful. I was looking at some data where stimuli are jittered but pretty closely spaced and the subjects had a lot of variability in response latency and duration. I found that I needed to include a long time interval for the CSPLIN model in order to achieve consistency, and at that point, it seemed like the model was capturing some of the signal from subsequent stimuli. When I try SPMG3 on these data, I get an almost identical iresp for the first 10 seconds, with a slightly higher peak in the SPMG3, so I’m thinking it might be preferable in this situation.

But I wish I understood exactly how you calculate the SPMG3 iresp from the three parameters for a given point in time post-stimulus-onset. Is that too complicated to describe?

Sally

I wish I understood exactly how you calculate the SPMG3 iresp from the three parameters for a given

point in time post-stimulus-onset. Is that too complicated to describe?

Not really. Assume that the three betas corresponding to the three SPM basis functions are b0, b1, and b2. The assembled response curve from the option -iresp in 3dDeconvolve is basically

b0 * G0(t) + b1 * G1(t) + b2 * G2(t),

where G0(t), G1(t), and G2(t) are the values of the three basis functions at the time point t. Does this make sense?

I guess I should have been able to figure that out! Thanks, Gang.