In this dissertation we propose a class of time series models for mixture data. We call these logistic mixtures. In such models the mixture's component densities have a generalized linear model (GLM) form. The regime probabilities are allowed to change over time and are modeled with a logistic...