Rémi Louf

Take prior predictive samples from the code example I shared, where do the values lie and what does this model _generate_? Truncation is about the generative model for your data,...

To generate censored data you generate from the full distribution (`normal(0,1)`) and then apply a `>= threshold` filter to your prior samples to mimic the _observation_ process. If that's not...

> Are you guys essentially talking about how the endpoints of the `clip`ed intervals are/should be handled/understood? Yes. In this PR are we talking about truncated random variables or random...

> Anyway, this PR was about truncation not censoring. You're right, will open a separate discussion for my remaining questions about censoring. This is still relevant and important here: >...

Does this means this PR is about RVs with a truncated probabilistic distribution then?

Ok, that's all I needed to know, and it's fine as long as it is clearly stated. I believe there is a difference between RVs with a truncated distribution (the...

As I expressed above, I have strong reservations when it comes to this PR and the meaning of the operator that is proposed. The problem can be seen clearly when...

Another example, which probably does not fail for the same reasons but illustrates how brittle the current framework is. The following returns, but the logprob is likely wrong as it...

I updated my original example to reflect the current API, and it still fails. The other example fails for a different reason so I masked the comments and opened a...

I'd be happy to have a look at this once I am done with NUTS in `aehmc`.