Suppose I want to do bayesian inference on the regression problem $\beta$ for Y = X$\beta$ + $\epsilon$ for $\epsilon_i \sim N(0,\sigma^2)$. The complication is that the prior for each component $\beta_i$ is a mixture of a point mass at 0 and a normal (I think this is called a spike and slab prior).

For example, $\beta_j \sim p\delta_{0} + (1-p)G$ for $G \sim N(0,\tau^2)$.

I want to write down the product of the prior and the likelihood but I find it difficult to do this with the dirac delta function in the prior. I tried introducing a latent variable to indicate which component of the prior I'm drawing from, but I wonder if there is a more principled way to write down the product of the likelihood and the prior.

I seem to have this problem generally whenever there is a dirac delta function in a likelihood or prior. Any general tips would be appreciated.



Your Answer

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

Browse other questions tagged or ask your own question.