Dear Marga,

Looking at the average over all time bins is not unreasonable, and as
you say, does not bias you towards any particular time point. However,
the downside is that you lose any temporal information (in that
contrast)—if you are averaging over all 13 bins, you could have just
used a single bin for the response. So, at some point it would
probably be worth considering whether you could glean additional
information from some aspect of the timing of the response.

Two other quick comments:

-If the neural activity of interest is linked to the onset of taste
percept, you may want to try to link the analysis to this event (it
wasn't clear to me what onset time you were using for your events);

-If the time between taste onset and swallowing is pretty consistent
across trials, it will be difficult to dissociate neural activity due
to these two events—if there is some variability in that delay, you
could potentially do so, and that might motivate modelling them as
separate events.

Hope this helps!

Best regards,

Jonathan

-- 
Dr. Jonathan Peelle
Department of Otolaryngology
Washington University School of Medicine
660 South Euclid, Box 8115
St. Louis, MO 63110
United States
http://jonathanpeelle.net/


On Mon, Oct 1, 2012 at 5:53 PM, Marga Veldhuizen
<[log in to unmask]> wrote:
> Dear SPMlist,
>
>
>
> In the study that I am currently analyzing we present flavors as small
> boluses of liquid to the subjects’ mouth while they are in the scanner. The
> flavors are cued by a visual stimulus (a picture of a bottle). Then the
> liquid is administered and the subject is instructed to swallow at the end
> of liquid administration.  This is the order of events:
>
>
>
> * at t = 0 s a visual of a beverage bottle with a specific color appears on
> screen
>
> * at t = 2.5 s, the liquid start dripping into the mouth (=taste percept
> onset)
>
> * at t = ~ 6.5 s, the subject swallows and exhales through the nose (=
> flavor/retronasal olfactory onset).
>
> * after a jitter of 4-11 s a rinse is presented
>
>
>
> The visual stimulus remains on until presentation of the rinse. Taste and
> retronasal perceptions may linger until rinse, but definitely decay after
> swallow.
>
>
>
> For this dataset I figured FIR modeling would be justified, as I didn’t want
> to impose a specific HRF. After all, some of our areas of interest may
> respond with one big long BOLD, and some may show a double BOLD (say
> starting at t = 0 and at t = 6.5). Indeed, this last type of response is
> exactly what I see in some of our areas when I look at PSTH for trials that
> contain a flavor versus trials with a flavorless baseline.
>
>
>
> My question:
>
>
>
> I’m currently creating contrasts at the second level. Does it make sense to
> use a t-contrast that looks at the average difference between flavor A and B
> across all of my 13 time bins (for example: repmat(-1, 1,13) repmat(1,
> 1,13))? This way I think I am not biasing towards an effect in any one bin
> over another, and it would be like looking at an area under the curve.
>
>
>
> I would very much appreciate your opinion on this!
>
>
>
> Many thanks,
>
>
>
> Marga Veldhuizen
>
>
>
>
>
> Maria Geraldine Veldhuizen, PhD
>
> Neurogastronomist
>
> Associate Research Scientist
>
> Affective Sensory Neuroscience
>
> The John B. Pierce Laboratory
>
> Yale University School of Medicine
>
> [log in to unmask]
>
> office: (203) 562 9901 ext. 251
>
> lab: (203) 562 9901 ext. 210
>
>