Compute psychometric functions based on d-primes?

1
Hi everyone,
this question isn't actually about the psychtoolbox, strictly
speaking. Still, I hope someone out there can help me.

I want to compare performance between several tasks. Some of these
tasks are 2AFC, others are yes/no. To have a common metric for these
tasks, my idea was to compute d-prime, since this is possible for
yes/no as well as for 2AFC. Another factor in this study will be
stimulus duration. Is it ok to compute psychometric functions based on
d-primes as a function of stimulus duration? And is it valid to
compare these psychometric functions (say, the estimated threshold)
between yes/no and 2AFC tasks?

Thanks a lot in advance,
Niko
2
On Sun, Oct 12, 2008 at 11:07 AM, niko.busch <niko.busch@...> wrote:

> And is it valid to
> compare these psychometric functions (say, the estimated threshold)
> between yes/no and 2AFC tasks?

Theoretically, there is a relationship between d' in yes/no and 2AFC
of sqrt(2) provided the assumptions of the equal-variance guassian
model hold for your data. There's an excellent explanation in the
Wickens text "Elementary Signal Detection Theory".

cheers,
-d

--
==========================
Donald Kalar, M.A., C.Phil.
Graduate Student,
Cognitive Neuroscience
UCLA Department of Psychology
1285 Franz Hall, Box 951563
Los Angeles, CA
90095-1563
==========================
3
Thank you Don for clarification! So, let's assume that the assumptions
for comparing d-prime for yes/no and 2AFC hold. Can I compute
psychometric functions based on these d-prime values? In fact, can
psychometric functions be computed based on any d-prime values at all?
Cheers,
Niko