endogeneity {endogeneity}

Recursive two-stage models to address endogeneity

Description

This package supports various recursive two-stage models to address the endogeneity issue. The details of the implemented models are discussed in Peng (2022). In a recursive two-stage model, the dependent variable of the first stage is also the endogenous variable of interest in the second stage. The endogeneity is captured by the correlation in the error terms of the two stages.

Recursive two-stage models can be used to address the endogeneity of treatment variables in observational study and the endogeneity of mediators in experiments.

The first-stage supports linear model, probit model, and Poisson lognormal model. The second-stage supports linear and probit models. These models can be used to address the endogeneity of continuous, binary, and count variables. When the endogenous variable is binary, it can be unobserved or partially unobserved, but the identification can be weak.

Functions

bilinear: recursive bivariate linear model

biprobit: recursive bivariate probit model

biprobit_latent: recursive bivariate probit model with latent first stage

biprobit_partial: recursive bivariate probit model with partially observed first stage

linear-probit: recursive linear-probit model

probit_linear: recursive probit-linear model

probit_linear_latent: recursive probit-linear model with latent first stage

probit_linear_partial: recursive probit-linear model with partially observed first stage

probit_linearRE: recursive probit-linearRE model in which the second stage is a panel linear model with random effects

pln: Poisson lognormal (PLN) model

pln_linear: recursive PLN-linear model

pln_probit: recursive PLN-probit model

References

Peng, Jing. (2023) Identification of Causal Mechanisms from Randomized Experiments: A Framework for Endogenous Mediation Analysis. Information Systems Research, 34(1):67-84. Available at https://doi.org/10.1287/isre.2022.1113