I have a psychophysics experiment and I am measuring whether participants can or cannot see the stimulus based on contrast. I have two options for my logistics regression. 1) use the raw data (0s and 1s) to indicate whether they did or did not see the stimulus. However, the paper I am basing my analysis on runs the binomial (probit) GLM on transformed data that takes into account false-positive rate. So option 2) is to follow that paper and have the outcome variable take values between 0 and 1. I then have many fewer data points because they get collapsed based on stimulus parameters to give the transformed outcome variable. So the question is: can I use the weights argument in R's GLM to specify how many trials are represented by each individual transformed data point? Sorry for the long explanation, but I thought some background would be relevant. I have already tried both options, as well as using the transformed outcome variable without weights, and they all yield different results.

Full article content could not be extracted automatically. Read the original below.