Statistics Made Easy 8: Causal Inferences and Endogeneity скачать в хорошем качестве

Statistics Made Easy 8: Causal Inferences and Endogeneity

In this video, we discuss endogeneity problems that can lead to biased estimates in regressions. We focus on three primary sources of endogeneity: selection bias, omitted variable bias, and reverse causality (or simultaneity). When endogeneity is present, making causal inferences can be challenging. In this case, it is often prudent to interpret the results as "associations" or "relationships" rather than causal effects. We also highlight some techniques to test for and address endogeneity using Stata. For additional background reading, I recommend these papers: 1) https://doi.org/10.1177/0149206320960533 2) https://doi.org/10.1177/0149206319868016 The Stata commands presented for handling endogeneity are as follows (you can install them using the Stata command "ssc install [command]"): 1) Propensity score matching: teffects psmatch 2) Coarsened exact matching (Iacus et al. 2012): cem 3) Lewbel (2012): ivreg2h 4) Oster (2019): psacalc 5) Sensitivity analysis (Cinelli & Hazlett, 2024): sensemakr 6) Impact Threshold of Confounding Variables (Busenbark et al. 2022): konfound Remember, the official Stata help files (type "help [command]" in the Stata command line) and the Stata forum (statalist.org) offer valuable support. Additionally, search engines (e.g., Google and Bing) and artificial intelligence tools (e.g., ChatGPT and Gemini) can assist you in finding the code you need. You can download the datasets, additional explanations, and the Stata "Do Files" for the tutorial by following this link: https://drive.google.com/drive/folder...