439 – Bayesian Modeling in Social Sciences
Bayesian Estimation with Flexible Priors for the Instrumental Variabls Models
Julianne Swenson
University of California at Santa Barbara
John Hsu
University of California at Santa Barbara
The instrumental variables method is a valuable tool in the analysis of simultaneous equations models. Since the estimation of coefficients in the model can be challenging, adept modeling of the covariance matrix is also important. The Inverse Wishart distribution is commonly used to provide a conjugate prior for the covariance matrix. However, the Inverse Wishart is limited in the flexibility to model prior information. We propose an alternative that alows for the specification of varying confidence levels for each element in the covariance matrix for the Instrumental Variables models. A real life example with corresponding results will be discussed in detail.