A novel two stage approach for causal effect estimation based on predictions of the response variable

Abstract

Research builds upon pre-existing knowledge by investigating the relationship between novel variables and variables whose inter-relationships are well understood. This scenario can often be described as a complex model of known structure incorporating latent and manifest variables. In parts of the model relating to well understood variables, parameter values might be well estimated as a result of previous research. For the rest of the model the parameters are unknown. The research question can be framed as the estimation of these unknown model parameters. Traditionally this would be add...