[visionlist] signal detection query

[Joseph Brooks]

I was afraid of something like this. We have thought about (and it has
been suggested by others in response to this message) just randomly
splitting the no-signal trials to create two separate estimates of FA.
However, I'm not convinced that this is really a solution as I expect
that the two randomly-selected samples are just drawn from the same
distribution. Thus, any difference between them should just be due to
noise and not due to any true differences in FA rate due to TMS
intensity. How could it be otherwise given that they are randomly
selected?!

It looks Greg may be correct...perhaps the best way to do this would
be to make it a block design (which is what we were trying to avoid)
and have separate testing sessions for the two TMS intensities and
thus have separate groups of no-signal trials for the different TMS
intensities. Any effect of TMS intensity on FA rate should then show
up as a between session difference.

Thanks for all the thoughts on this!

Joe



On 21 July 2010 01:37, Dr G J Davis <gjd1000 at cam.ac.uk> wrote:
> Dear Joseph,
> Having no independent assessment of FA rate for each condition must, I
> think, limit interpretation.  Without it you cannot determine whether any
> difference between TMS conditions in P(Hit) reflects a change in sensitivity
> or alternatively in criterion.
> If the distinction is not important for your purposes, you need not use SDT.
> If it is important, maybe separate blocks of high v low TMS would be best as
> they would generate separate estimates of P(FA) for each condition. Is that
> feasible in this case?
> Dr Greg Davis, Lecturer
> Dept of Experimental Psychology
> University of Cambridge
> Downing Street, Cambridge, UK
> Dean and Fellow in Psychology
> Murray Edwards College
> Huntingdon Rd, Cambridge, UK
> www.psychol.cam.ac.uk/viscog
> On 20 Jul 2010, at 21:03, "Todd S. Horowitz" <toddh at search.bwh.harvard.edu>
> wrote:
>
> Hi Joe
> (1) As far as I know, the absolute numbers of trials in each cell is not
> really relevant. What might happen is that having more old pictures will
> bias the subjects to respond "old", but that will show up in the
> criterion/bias measures. I don't think that it affects the underlying
> assumptions of the model.
> (2) Using the same false alarm rate is commonly done in similar
> circumstances. Again, I don't think it poses a problem, as long as you are
> mixing the two types of old pictures in the same block of test trials.
> thanks
> Todd
> On Jul 20, 2010, at 1:03 PM, Joseph Brooks wrote:
>
> We are calculating d-prime for a visual memory experiment. We have two
> issues which are unclear to us and we are hoping that someone may be
> able to give us some guidance or direct us to the best resource.
>
> In our experiment people see a set of pictures. Then later they are
> presented with a set of pictures (the test set) and asked whether they
> saw each picture before (old/new judgment). The test set includes both
> the previously seen pictures (2/3 of the total test set) and some
> novel pictures that were not seen previously (1/3 of the total test
> set). Unfortunately we had to have this uneven number of old and novel
> pictures because of constraints on our stimulus set.
>
> (1) Does the calculation of d-prime need to be adjusted in any way
> because of greater number of signal (old pictures) than noise (novel
> pictures) trials?
>
> (2) Half of the previously seen pictures were seen during high
> intensity transcranial magnetic stimulation whereas the other half
> were seen during low intensity stimulation. We will calculate the hit
> rate separately for the high and low conditions. However, none of the
> novel stimuli (which will be used to calculate the false alarm rate)
> had any stimulation at all. Thus, the two hit rates mentioned above
> don't have two separate corresponding false alarm rates. Rather there
> is just one false alarm rate based on all of the novel trials. Any
> d-prime values that we calculate from the two hit rates and the one
> false alarm rate will share the same false alarm rate. I don't see any
> other way to do this and I'm assuming that this does NOT pose any
> problems down the line for comparing the two d-primes (given that they
> are now no longer independent). Am I wrong here?
>
> Any help with these issues will be much appreciated!
>
> Regards,
>
> Joseph Brooks
> Institute of Cognitive Neuroscience
> University College London
> _______________________________________________
> visionlist mailing list
> visionlist at visionscience.com
> http://visionscience.com/mailman/listinfo/visionlist
>
> Todd S. Horowitz, PhD
> Assistant Professor of Ophthalmology
> Harvard Medical School
> Associate Director
> Visual Attention Lab
> Brigham & Women's Hospital
> 64 Sidney Street, Suite 170
> Cambridge, MA 02139
> phone:  (617) 768-8813
> fax:    (617) 768-8816
> http://search.bwh.harvard.edu/
>
>
> The information in this e-mail is intended only for the person to whom it is
> addressed. If you believe this e-mail was sent to you in error and the
> e-mail
> contains patient information, please contact the Partners Compliance
> HelpLine at
> http://www.partners.org/complianceline . If the e-mail was sent to you in
> error
> but does not contain patient information, please contact the sender and
> properly
> dispose of the e-mail.
>
> _______________________________________________
> visionlist mailing list
> visionlist at visionscience.com
> http://visionscience.com/mailman/listinfo/visionlist
>