ABSTRACT
This chapter describes the problems that can arise from treating such binary outcome variables as if they are continuous. It considers two possible implementations: logistic growth curve analysis (GCA) and quasi-logistic GCA using empirical logits. In the psychological and neural sciences we typically treat outcome variables as continuous – as if they could have any value. Logistic regression models the binomial process that produces binary data, so the outcome variable in the data set needs to be those binary data. As with the linear GCA on fixation proportions, the empirical logit GCA can handle the full random effects structure, the computation is faster, and the results do not include p-values. Logistic GCA will use the numerator and denominator from computing that proportion: N is the number of trials for each participant in each condition and sumFix is the number of trials on which the target was fixated by each participant in each condition in each time bin.
