3dLSS with 3ddeconvolve TENT function

AFNI Message Board
1

Hello! I'm running 3ddeconvolve with stim_times_IM for one of my regressors from which I am aiming to receive a deconvolved time series of beta values per stimulus time. Because I would like this to be deconvolved rather than convolved, I used the TENT function.

The data are naturalistic, thus a fast event-related design and highly correlated. I was considering running 3dLSS to account for this, but I'm wondering if the pull-one-out approach will be per beta value (not ideal since I'm receiving 15 beta values per stimulus time), or per stimulus input time (aka pull out 15 betas at a time based on the original 3ddeconvolve's 1D timing file)?

If it will pull out each beta, do you have another recommendation for how I might account for the highly correlated nature of my fast event-related naturalistic design, while still using a deconvolution instead of a convolution? Thank you!

2

Assume that the "stimuli" in your naturalistic data occur close to each other (e.g., within a few seconds). In that case, estimating the BOLD response at the stimulus level while simultaneously extracting its shape becomes quite challenging.

With 3dLSS, you may need to assume a fixed HRF. A more flexible alternative is to use GLMsingle, which allows for more adaptable HRF estimation.

Gang Chen

3

Thanks, Gang. We will give GLMsingle a shot!