Alice,
I assume you have an event-related design. In other words, each stimulus shown on the screen to the subject lasts for one TR or less. So you presume your HRF would last for 20 seconds? Maybe 15 (or even 10) seconds would be good enough? Let's go with 15 s, and set b=0 and c=15. You don't have to, but choosing n = 7 would span each TENT nicely over 4 TR's so that the beta corresponding each TENT would be exactly the BOLD response at every other TR grid.
> The main thing that confuses me about the results I got from the
> deconvolution after using 8 tent functions were the graphs. What am I
> supposed to do with the 8 graphs? According to slides 5 and 6 on the pdf-- am
> I supposed to add these 8 graphs together to get a final HRF response?
The modeled HRF through deconvolution is the linear combination of those n tents, and those beta's are supposed to be coefficients for the linear combination. More specifically, those beta's are actually the BOLD response at some grids. So if you specify the number of tents carefully, those beta's would be exactly the BOLD response at some TR grids.
In other words, if you plot out those sequential beta's as a time series, it would be your modeled HRF for each stimulus type, lasting for whatever time length you set in TENT(b,c,n), which is c-b. This has nothing to do with your total time points in the data.
HTH,
Gang