Custom psychometric functions

[Duls91]

customPFDemos.zip
Hello,

I wanted to implement two custom psychometric functions and test their performance with the simple routine from qpQuestPlusCoreFunctionDemo.m.
Attached are the custom PFs "Gumbel_pal" and "Weibull_pal". In the respective qpQuestPlusCoreFunctionDemo_Gumbel and qpQuestPlusCoreFunctionDemo_Weibull scripts, I specified the PFs and adjusted the stimulus domains.

The idea behind it:
The original Weibull function in mQuestPlus is a log-Weibull (or Gumbel) adjusted for the dB scale (therefore dividing by 20).
I wanted to write a more general log-scale Weibull ("Gumbel_pal").
Also, on a linear scale, a conventional Weibull ("Weibull_pal") would be useful.

Testing these two PFs leads to bad performance though. The stimuli presented by Quest seem to be way off and the parameter estimates for threshold or slope are disastrous.
I am not sure whether I adapted the scripts at all places to work with the custom PFs. Plotting the functions seems to result in a valid output i.e.

stimFine = linspace(-60,-20,100)';
y = qpPFGumbel_pal(stimFine, [-28, 0.15, 0.01, 0.01]);
plot(stimFine, y(:,2), 'linewidth', 2);

Hope someone can resolve this issue with me. Thank you!

[DavidBrainard]

Dear Armin, Unfortunately, my current schedule will prevent me from being able to look at this for at least a few months. I do note that someone else already contributed a qpStandardWeibull function to mQUESTPlus. I don't immediately see a demo for that, but perhaps it will be helpful to you. If you do figure this out, please send updated code and demos and I will sooner or later review and add to the contributed folder. If you can't figure it out by, say, mid-March, post back and I will do my best to look at it. Best, David From: Armin Drusko <notifications@github.com> Reply-To: BrainardLab/mQUESTPlus <reply@reply.github.com> Date: Friday, January 31, 2020 at 8:01 AM To: BrainardLab/mQUESTPlus <mQUESTPlus@noreply.github.com> Cc: Subscribed <subscribed@noreply.github.com> Subject: [BrainardLab/mQUESTPlus] Custom (#24) customPFDemos.zip<https://github.com/BrainardLab/mQUESTPlus/files/4139424/customPFDemos.zip> Hello, I wanted to implement two custom psychometric functions and test their performance with the simple routine from qpQuestPlusCoreFunctionDemo.m. Attached are the custom PFs "Gumbel_pal" and "Weibull_pal". In the respective qpQuestPlusCoreFunctionDemo_Gumbel and qpQuestPlusCoreFunctionDemo_Weibull scripts, I specified the PFs and adjusted the stimulus domains. The idea behind it: The original Weibull function in mQuestPlus is a log-Weibull (or Gumbel) adjusted for the dB scale (therefore dividing by 20). I wanted to write a more general log-scale Weibull ("Gumbel_pal"). Also, on a linear scale, a conventional Weibull ("Weibull_pal") would be useful. Testing these two PFs leads to bad performance though. The stimuli presented by Quest seem to be way off and the parameter estimates for threshold or slope are disastrous. I am not sure whether I adapted the scripts at all places to work with the custom PFs. Plotting the functions seems to result in a valid output i.e. stimFine = linspace(-60,-20,100)'; y = qpPFGumbel_pal(stimFine, [-28, 0.15, 0.01, 0.01]); plot(stimFine, y(:,2), 'linewidth', 2); Hope someone can resolve this issue with me. Thank you! — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#24?email_source=notifications&email_token=AAI24ZIQIDXC4BFXMQ25U73RAQOJXA5CNFSM4KOGVN62YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IKEG5DA>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AAI24ZIWKDEJJCPVZTG346LRAQOJXANCNFSM4KOGVN6Q>.

[Duls91]

Dear David,

thank you for your reply. I did miss the already contributed Weibull function. Still, even testing my Demo with this PF is not leading to a good performance.

Maybe @aernesto could help me out here. I would appreciate your thoughts on this.

The figures below show the PFs after questPlus threshold estimation with the Weibull and contributed StandardWeibull respectively.

[Duls91]

Dear David,
Unfortunately, I haven't been able to solve this issue yet.

Could you have a look at it?

Best,
Armin