Study on whether marriage affects depression based on causal inference

Abstract

This paper applies causal-based machine learning algorithms to evaluate the causal effect of marriage on depression. The paper verifies the reinforcement of adopting causal inference through the relationship between causality and correlation and confounding bias and selection bias. In this paper, we firstly implement meta learner to estimate and analyse the causal effects. Considering the influence of confounding factors, we utilize two stages of least squares estimation and deep IV estimation based on instrumental variables to fully evaluate the causal effects. The evaluation of linear and nonlinear models shows different results, which is worthy of discussion in future studies. In conclusion, people in the rural region who get married are slightly less likely to get depressed in the future.