In the book "Flexible Imputation of Missing Data" by Van Buuren, the following algorithm is presented

enter image description here

I think I understand the algorithm as given, but I would like to know what is the "more elaborate modeling" needed to "define an explicit Bayesian method for drawing $\beta$ and $V$ from their exact posteriors". How could the algorithm be extended to achieve this ?

  • $\begingroup$ I believe they may be referring to the Multivariate Probit Model. $\endgroup$
    – whuber
    Oct 29, 2014 at 18:11
  • $\begingroup$ @whuber what makes you believe so ? How does it enable drawing $\beta$ and $V$ from their exact posteriors, and how would the algorithm be amended to do so ? $\endgroup$
    – Joe King
    Oct 29, 2014 at 18:20
  • $\begingroup$ Please read the references. An accessible one, which shows up as second in my Google search, is Aline Tabet's MS thesis. $\endgroup$
    – whuber
    Oct 29, 2014 at 18:26

1 Answer 1


To draw $\beta$ and $V$ from their posteriors, you would need to define a prior over them. This is the "more elaborate modeling" that they are trying to avoid. Once you have chosen priors, it is pretty straightforward (using MCMC) to draw samples from the posterior.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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