Sliding quest

1
Hi all,
I have a Quest question.
I use it to monitor my subject's perceptual threshold. With my current settings, the Quest structure takes all previous trials into account.
I would like to create a sort of 'sliding Quest' that would take only the previous (say) 50 trials and follow my subjects' perceptual threshold more closely. Is it possible?
Thanks for any advice.
Best,
Max
2
Thank you Genna,

I think I got your point. Now if I believe that the threshold varies over time, does it make sense to remove trials that are the furthest in time from the current trial? That way, I will follow the actual threshold with more accuracy, won't I? Or is it just not the way to go to follow a varying threshold?
See for example this plot.
Thanks for any advice,
Max

2011/8/31 gennaer <gennaer@...>

Max,

I think the easiest would be to, starting on the 51st trial (after you get the response and run QuestUpdate):
-remove the first intensity and response values from the q structure (better to save them somewhere)
-change the trialCount from 51 to 50 since this is what QuestRecompute loops over to recreate the pdf from the response history (note that now you can't use trialCount as a counter to know when to end your study)
-generate a new q structure by running QuestRecompute to regenerate the pdf from the modified history
-use this new q structure in computing the testing value for the subsequent trial (e.g. in QuestQuantile)

Note: if you use QuestMode to generate tTest, it might not be any different than if you used all of the values.

However, I'm not really sure you'd get significantly different results if you store 50 trials instead of all of them and I don't think it'll really help. One of the advantages of QUEST is supposed to be precisely that it can take all trials into account. And while the initial trials are the ones that are most likely to test intensities that are furthest from the true threshold, taking them out one trial at a time will change the pdf by a virtually imperceptible amount. As a result, your tTest won't change that much, especially for later trials.

Genna



--- In psychtoolbox@yahoogroups.com, "maximilienchaumon" <maximilien.chaumon@...> wrote:
>
> Hi all,
> I have a Quest question.
> I use it to monitor my subject's perceptual threshold. With my current settings, the Quest structure takes all previous trials into account.
> I would like to create a sort of 'sliding Quest' that would take only the previous (say) 50 trials and follow my subjects' perceptual threshold more closely. Is it possible?
> Thanks for any advice.
> Best,
> Max
>