Hi everybody.
I ran an experiment on memory and calculated the percentage of hits as usually reported in the literature. However, the reviewer asked me to transform my hits into d-prime values. I was able to calculate the d-prime values using the psyphy library in R. But, in my original analysis, I calculated the percentage of hits for each individual, then the average value and the confidence interval of the mean. My question is if I can calculate the mean and the confidence intervals of the d-prime values of my individuals, or if I have to calculate the d-prime value for all the subjects? Do you know if there is a measure of dispersion for the d-prime values?
Thanks for your help
It depends on the size of individual data. Since hits (and false alarme and other confusion cells) are estimates of conditional probabilities, they may not be accurate when the data size is small. Macmillan & Kaplan (1985, Psych.Bull) discuss pros and cons of computing d' from grouped data.
Gourevitch & Galanter (1967, Psychometrica) developed a method to compute variance of d', which can be used to compute confidence intervals. We used this method to compare individual d's (Ward & Mori, 1996, JASA).
- Badges
- Science topic