I have several miscellaneous questions about QUEST.
I am trying to use QUEST in a paradigm that does not have a fixed nAFC structure.
n changes between 2 to 6 on every trial. This posesses the problem of assigning a value to gamma.
My solution was to have a weigted avarage of all probabilities based on how many times they would occur and use this value for gamma. Is this a sensible choice to make or do you have any suggestions on how to decide what to assign to gamma in this case?
I don’t want to have a fixed number of trials. I want to stop my experiment when the variance of my posterior reaches a certain point. Can I use the value from QuestSd(q) as a criteria to stop my experimental while loop for this purpose?
Thank you in advance!
Andrew “Beau” Watson and I created the original Quest (using maximum likelihood) which we published in 1981, I think. Later I created the MATLAB implementation, which is Bayesian.
(2. Yes you can use QuestSD in your stopping rule. Beau and I played with various stopping rules. We thought this would be great, but in fact it tends to result in wasting a lot of trials on runs that went badly (eg. the participant was distracted for a few trials). Also, participants like to know how many trials remain. But try it and see how it works for you.
(1. Hmm. In principle Quest would allow you to update your posterior PDF using a different psychometric function on every trial. This is the first time I’ve considered this case, so it wasn’t envisioned in the current Quest implementation, but I’d guess it could easily be achieved. It might require minor modification of the current code.
Thank you for your feedback, Denis!