- A model with binary dependent variables based on the logistic distribution. Its use is normally associated with the maximum likelihood estimation procedure instead of ordinary least squares. The idea is to posit an underlying continuous latent variable, such that, given binary date, whether the score is 0 or 1 depends upon a critical value of the latent variable. Thus, if the binary variable is sick/well, the critical threshold might be a stipulated value of a quality-adjusted life-year. The latent variable is then posited to be a linear function of other variables including unobservable ones and the error term is assumed to be standard logistically distributed. Introduced by Berkson (1944).