George C. Tiao

Overview

George C. Tiao (1933–) is a Bayesian statistician, formerly University of Wisconsin (1960s) and later W. Allen Wallis Professor at the University of Chicago Booth School of Business. He co-authored with George E.P. Box the foundational text Bayesian Inference in Statistical Analysis (Addison-Wesley, 1973), which systematized Bayesian analysis for the normal linear model, outlier detection, and time series. His 1964 collaboration with Arnold Zellner introduced the "double-t" distribution as the Bayesian posterior for regression coefficients under independent heterogeneous error variances, and his work with Box on intervention analysis (1975) formalized the statistical treatment of structural breaks in time series.

Key Contributions / Features

• Tiao-Zellner (1964): Joint posterior of β from two regression samples with independent unknown variances $\sigma_1^2$, $\sigma_2^2$ is a product of two multivariate t-distributions (the "double-t"). Developed an asymptotic expansion in $\nu_1^{-1}$ and $\nu_2^{-1}$ to make this tractable. Established that Theil's (1963) mixed estimator is the Bayesian limit as $\nu_1 \to \infty$ with $\sigma_1^2$ known. Biometrika 51(1/2): 219–230.
• Box-Tiao (1973): Bayesian Inference in Statistical Analysis — comprehensive Bayesian treatment of regression, outliers, and time series under noninformative priors; the standard reference for Bayesian econometrics alongside Zellner (1971).
• Box-Tiao (1975): Intervention analysis for time series — transferred the Bayesian change-point framework to ARIMA models and business/economics applications.

Related

• Bayesian Linear Regression
• Seemingly Unrelated Regression
• Arnold Zellner
• Tiao-Zellner (1964)