In behavior modification studies dynamically changing factors have complex effects on subsequent actions. The question is how to accommodate baseline and post-baseline factors in one flexible model and assess effects on outcomes that are of clinical relevance. For example, in a longitudinal study about smoking relapse, potential predictors for successful smoking cessation include: nicotine dependence, mood disturbance, and weight change. Several models could be fit to the data depending on the clinical question of interest; often in exploratory analyses, however, several models are fit to the data. This presents a particular challenge to the analyst if competing inferences are made. We developed a single flexible model that can be used to test a variety of null hypotheses, thus providing the analyst a more efficient tool for evaluating complex time-varying covariates.