Semiparametric Bayes instrumental variable estimation with many weak instruments

We develop a new semiparametric Bayes instrumental variables estimation method. employ the form of regression function reduced-form equation and disturbances are modelled nonparametrically to achieve better preditive power endogenous variables, whereas we use parametric formulation in structural equation, which is interest inference. Our simulation studies show that under small sample size proposed method obtains more e? cient estimates very precise credible intervals compared with existing IV methods. The methods fail reject null hypothesis higher probability, due larger variance estimators. Moreover, mean squared error may be less than 1/30 procedures even presence weak instruments. applied our Mendelian randomization dataset where large number instruments available specification appropriate. This instrument case; hence, non-Bayesian approach yields inefficient estimates. obtained statistically significant results cannot by methods, including standard Bayesian IV.