A two-part random effects model for semicontinuous longitudinal data

Joe Schafer

Penn State University

Refreshments: 3:30 - 4:00 p.m. Friday, November 12, at 327 Yost
Talk: 4:00 - 5:00 p.m. Friday, November 12, at 327 Yost.

A semicontinuous variable has a portion of responses equal to a single value (typically zero) and a continuous, often skewed, distribution of values among the remaining responses. In econometric analyses, variables of this type have been described by a pair of regression models, a logistic model for the probability of nonzero response and a conditional linear model for the mean response given that it is nonzero. In this paper, we extend two-part regression to longitudinal settings by introducing random coefficients into both the logistic and the linear parts. Fitting this two-part random-effects model poses computational challenges similar to those found in generalized linear mixed models. Maximum-likelihood estimates for the fixed coefficients and variance components are found by an approximate Fisher scoring procedure based on high-order Laplace approximations. The technique is illustrated with data from the Adolescent Alcohol Prevention Trial, where we examine reported recent alcohol use for students from the 7th to the 11th grade and its relationships to parental monitoring and rebelliousness. This is joint work with Maren K. Olsen of Penn State.

Questions? Nidhan Choudhuri