Quest problems

1
Through both empirical data I'm collecting and a number of simulations, I've discovered
that the resulting threshold returned by the Quest function is greatly biased by the initial
threshold guess I give the function (tGuess). That is, the lower the initial threshold
estimate, the lower the threshold returned by the function. I'm using the recommended
values for tGuessSD (3), range (5), beta, delta, and gamma, and the program is using a
termination rule (determined by the size of the confidence interval).

Has anybody encountered this problem, and if so, how can I avoid biasing my resulting
estimates? Is there any literature on the implementation of Quest?

Thank you in advance for your help--
Sara
2
dear sara

yes, i echo david's comment. formally, you have to expect an effect of
the prior. in a bayesian framework the posterior pdf (which i believe
you're using to make your final estimate) is the product of the prior
and the likelihood. so we expect an effect, the question is how much.
for that we need to see your code and a quantitative summary of what
you found.

that code has been critically examined and tested many times, so we'll
be surprised if it's doing other than claimed, but even nasa finds
bugs, so perhaps there's one here too. please send details.

best

denis



On Apr 20, 2005, at 8:37 AM, David Jones wrote:

>
>
> Hi Sara,
>
> Can you post the code you used in your Quest simulations?
> This would help us diagnose what is going on.
> Quest is mature and stable, so I would be surprised
> if it is that sensitive to the initial guess.
>
> -- David Jones
>
>
> On Apr 19, 2005, at 10:29 PM, sara4417 wrote:
>
>>
>>
>> Through both empirical data I'm collecting and a number of
>> simulations, I've discovered
>> that the resulting threshold returned by the Quest function is greatly
>> biased by the initial
>> threshold guess I give the function (tGuess). That is, the lower the
>> initial threshold
>> estimate, the lower the threshold returned by the function. I'm using
>> the recommended
>> values for tGuessSD (3), range (5), beta, delta, and gamma, and the
>> program is using a
>> termination rule (determined by the size of the confidence interval).
>>
>> Has anybody encountered this problem, and if so, how can I avoid
>> biasing my resulting
>> estimates? Is there any literature on the implementation of Quest?
>>
>> Thank you in advance for your help--
>> Sara
>
>
>
>
>
>
> Post your message to: psychtoolbox@yahoogroups.com
> Please indicate OS9, OSX, or WIN version, and include your full name.
> Denis Pelli, David Brainard, and Allen Ingling.
> http://psychtoolbox.org
>
> Yahoo! Groups Links
>
>
>
>
>
>
>