Using {projpred} latent projection with {brms} Weibull family models

[rtnliqry]

@fweber144

Yes, this is what I meant with … However, as noted above, this more general (and hence more desirable) solution is probably more tedious to implement.

Got you, sorry, I was slow to realise this!

@avehtari

Now I think we need to sample the missing censored observations from a truncated Weibull with the left truncation at the censoring time. Given the augmented observations, we can then use just the latent of Weibull (we can’t mix latent of Weibull and latent of Bernoulli).

@avehtari Did you have any thoughts on this?