This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: For inference about , a sample from the marginal posterior, p( data) would suffice. For inference about , a sample from the marginal p β β β σ σ σ σ ′ Σ σ × = σ 2 2 2 2 1 2 2 i i osterior of , p( data) would suffice. Can we deduce these? For this problem, we do have conditionals: p(  ,data) = N[ , ( ) ] (y ) p(  ,data) = K a gamma distributio i 2 b X'X β x β β σ 2 n Can we use this information to sample from p( data) and p( data)? β Part 24: Bayesian Estimation The Gibbs Sampler p Target: Sample from marginals of f(x1, x2) = joint distribution p Joint distribution is unknown or it is not possible to sample from the joint distribution. p Assumed: f(x1x2) and f(x2x1) both known and samples can be drawn from both. p Gibbs sampling: Obtain one draw from x1,x2 by many cycles between x1 x2 and x2x1. n Start x1,0 anywhere in the right range. n Draw x2,0 from x2x1,0. n Return to x1,1 from x1x2,0 and so on. n Several thousand cycles produces the draws n Discard the first several thousand to avoid initial conditions. (Burn in) p Average the draws to estimate the marginal means. ˜˜™™™ ™ 9/34 Part 24: Bayesian Estimation Bivariate Normal Sampling ˜˜™™™ ™ 10/34 ρ ÷ ÷ ρ = Γ ÷ ÷ ÷ Γ θ θ 1 1 1 2 2 2 r r 1 2 1 Draw a random sample from bivariate normal , 1 v u u (1) Direct approach: where are two v u u 1 independent standard normal draws (easy) and = ÷ ρ ΓΓ θ = ρ θ = ρ ÷ ρ ρ ρ ρ ρ 2 1 2 2 1 2 2 2 2 1 1 1 such that '= . , 1 . 1 (2) Gibbs sampler: v  v ~ N v , 1 v  v ~ N v , 1 Part 24: Bayesian Estimation Gibbs Sampling for the Linear Regression Model ˜˜™™™ ™ 11/34 σ σ ′ Σ σ × σ = i 2 b X'X β x β β 2 2 1 2 2 i i p(  ,data) = N[ , ( ) ] (y ) p(  ,data) = K a gamma distribution Iterate back and forth between these two distributions Part 24: Bayesian Estimation Application – the Probit Model ˜˜™™ ™ 12/34 ′ = ε ε = i i x + β β i i i i i i (a) y * ~ N[0,1] (b) y 1 if y * > 0, 0 otherwise Consider estimation of and y * (data augmentation) (1) If y* were observed, this would be a linear regression (y would not be useful ′ i  β β x β i i i i since it is just sgn(y * ).) We saw in the linear model before, p( y * , y ) (2) If (only) were observed, y * would be a draw from the normal distribution with mean and variance 1....
View
Full Document
 Fall '10
 H.Bierens
 Econometrics, Normal Distribution, Maximum likelihood, Estimation theory, BAYESIAN ESTIMATION, Bayesian statistics

Click to edit the document details