Text has a long history in psychological and education research, but appropriate quantitative methods are surprisingly underdeveloped. In this talk, I introduce a latent variable measurement modeling framework for text, topic modeling, that has been recently applied in psychology and education. Researchers have tried to link these latent variables (topics) from text to other measures in a regression framework using a two-stage approach. Two-stage approaches with latent variable models, however, can be problematic. I will discuss a new model I developed that addresses this research gap by extending supervised topic modeling to incorporate covariates and latent topics to model an outcome. I will describe a Bayesian estimation algorithm and R package to fit the proposed model. I will also present results from a simulation study evaluating the two-stage approach and the proposed model. Finally, I will demonstrate this approach in a study of emotional dysregulation and interpersonal conflict in the context of nonsuicidal self-injury.