I'm reading the Horseshoe prior regression paper which formulates regression like so:

$(y|\beta) \sim N(\beta, \sigma^2I)$

$(\beta_i|\lambda_i,\tau) \sim N(0, \lambda_i^2 \tau^2)$

$\lambda_i \sim C^{+}(0,1)$

where $C^{+}(0,1)$ is the half-Cauchy distribution.

In section 2.1 of the paper it says:

$E[\beta_i | y_i, \lambda_i^2] = (\frac{\lambda_i^2}{1 + \lambda_i^2}) y_i + (\frac{1}{1 + \lambda_i})0$

Where does this expression come from?

  • $\begingroup$ This conditional is Normal, as a conjugate prior should be. $\endgroup$ – Xi'an Dec 10 '17 at 22:21

By the usual conjugacy of the normal prior and likelihood, the posterior distribution (note also that $\tau$ and $\sigma$ are fixed to 1) is $\beta_i|y_i, \lambda_i\sim N(\lambda_i(\lambda_i+1)^{-1}y_i, \lambda_i(\lambda_i+1)^{-1})$. And so the expectation they show is simply the expectation of this posterior.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.