 |
 |
I am sending this to the list for a colleague.... Does it make sense to introduce sample weights into a Bayesian analysis of survey data, given that we know (or assume) how the parameters are distributed? If we do this, then should we apply it to the likelihood portion of the analysis? Also, would the problem with *not* including survey weights be exacerbated if some of the weights are really low and some weights are really high (i.e.., the actual sample is not random without the weighting)? Finally, any references on Bayesian analysis of survey data are greatly appreciated. Best wishes, Wendy Martinez