dissertation help please...signal detection and d-prime alternative?

#1
Hi!

Ive just conducted my analysis for my dissertation and im worried i have done it completely wrong. this is the basic design:
2 stages; 1. encoding - participants viewed stimuli while recording pupil responses
2. recognition - stimuli shown from encoding mixed in with new stimuli, participants able to choose from 5 options of recognition; 3 levels of familiarity (strong, moderate, weak), recollection and novetly. Have been told to calculate signal detection, so have worked out the hits and false alarms for all 3 levels of familiarity and recollection responses and then misses and correct rejections for novelty responses.
But was told that instead of calculating dprime, an easier method is to minus the proportions of false alarms from the proportion of hits, so for every level of familiarity (and recollection responses) I divided the total number of responses that were hits by the total number of trials. Then I converted these calculations for each level of familiarity into t-tests and ANOVAs in spss.
So all was fine, until I realised that while working out signal detection and proportions in excel, when there was no response (for example, many participants didnt give a recollection response), I simply put a 0 in the appropriate cell. However I have now realised that this affects my overall averages and consequently my analysis in spss, as oppose to leaving the cell blank. Particularly in spss, leaving the cell blank or putting a 0 significantly affects my overall analysis, so much so, that one route causes the findings to be significant, and the other route does not!
Which is the correct route to take? Do i use 0 for no response or simply leave it blank, especially for calculating proportion averages in both excel and spss?
Furthermore, for reaction time and pupil dilation calculations, does putting 0 for no response (as if they didnt give a recollection response, there was no reaction time) make a difference to my total analysis?

Help would be much appreciated? Supervisor has chosen to go away for the easter hols and Im supposed to be handing in a first draft in 3 weeks. Without working out this problem, I cant really interpret my findings and continue with my writeup?:(

many thanks

Lousie
 
#2
If the person did not respond you want to leave the cell blank. If, for example, you put zero as RT excel/SPSS will calculate reaction time analysis assuming that those individuals took 0 ms to respond (ie were very very fast). This is going to lower your mean RT for any condition that has a greater number of missing responses.

By the way in case it comes up again this formula can be used to easily calculate d prime scores in excel:
=NORMINV(hit-rate,0,1) - NORMINV(false-alarm-rate,0,1)
 
#3
Hi there,

I'm using exactly the same measurement as you in my dissertation. I tried to do it your way but had the same problem. BUT if you use d' prime (which is easy to calculate using that formula in excel) you still get the problem of 0 or 1 scores. The way to fix this is as follows:

Calculate a minimum and maximum score for your data.

It is possible that a respondent detects every signal, giving a hit rate of H = 1.00, or a respondent might make no false alarms giving FA = 0. A right-tail p value of 0 corresponds to z = infinity while a p value of 1 corresponds to z =negative infinity. This poses a problem for computing d’!

A conventional adjustment is to set the minimum p = 1/N where N is the number of trials used in the calculation of p. We set the maximum value for p = (N-1)/N.

http://wise.cgu.edu/sdtmod/measures3.asp
This online tutorial answers everything. Hope this isn't too late!!

Caz