[pymvpa] BOLD delay

[Christopher J Markiewicz]

On 02/11/2016 08:28 AM, Salim Al-wasity wrote:
> Dears
> When we do the classification, do we need to take into our account the
> delay of the BOLD signal and shift our samples by 5 sec.?
> 
> Sincerely
> Salim Al-Wasity
>     PhD student
>     University of Glasgow

Generally speaking, you should model the the events of interest with
hemodynamic response functions and use the beta estimates as your
samples for classification. If you've done the HRF modeling, then
there's no need to do an additional shift.

If you've used a sparse-acquisition paradigm, and HRF modeling isn't
feasible (though I would consider the following before ruling it out:
https://www.ncbi.nlm.nih.gov/pubmed/23616742), then, yes: You want to
label your volumes based on the peak BOLD response.

Hope this helps.

-- 
Christopher J Markiewicz
Ph.D. Candidate, Quantitative Neuroscience Laboratory
Boston University