Estimating causal effects in linear regression models with observational data: The instrumental variables regression...

Area of Science:

• Psychology
• Econometrics
• Statistical Modeling

Background:

• Causal inference in psychology often faces challenges with confounding variables.
• Instrumental variable methods offer a robust approach to address endogeneity.
• Instrumental variable regression (IVR) is an underutilized technique in psychological research.

Purpose of the Study:

• To provide a tutorial on applying instrumental variable regression (IVR) within a structural equation modeling (SEM) framework.
• To demonstrate how to identify and select optimal instruments for IVR.
• To guide researchers in drawing stronger causal inferences in psychological studies.

Main Methods:

• Utilized instrumental variable regression (IVR) by incorporating predictors of predictors (instruments).
• Employed structural equation modeling (SEM) with Maximum Likelihood (ML) estimation.
• Developed and provided code for instrument identification, subset selection, and SEM application.
• Demonstrated estimation using traditional econometric methods like 2-stage least squares (2SLS).

Main Results:

• IVR models can be reliably estimated and tested within SEM frameworks.
• The proposed methods facilitate the identification and selection of valid instruments.
• 2-stage least squares regression is shown to be a multistage SEM estimator for IVR.

Conclusions:

• Instrumental variable methods, particularly IVR within SEM, are powerful tools for enhancing causal inference in psychology.
• Researchers are guided on practical implementation, including instrument selection and SEM application.
• The study promotes wider adoption of IVR for more rigorous causal claims in psychological research.