kgoldfeld

No worries. The code is a little involved, but I'll include here. There is a [blog post](https://www.rdatagen.net/post/2021-12-21-controling-type-1-error-rates-in-rcts-with-interim-looks-a-bayesian-perspective/) describing what I doing, and includes all of the code. ```r bayes_fit m_effect...

So, I've done a little benchmarking with a simpler example (see at the end for the R code), and the results are indeed curious. When run with 4 parallel chains...

I guess no single package is ever perfect, and there is never enough time to make it so (unless you are Hadley Wickham and get paid to do it). I...

Just to be clear, I am not using `brms`, but coding directly in `Stan`. But, that is not really relevant. Yes, each case is within one execution of `bayes_fit`. As...

Yes - I have been considering each line of my data frame (using apply) or each element of the list (using MAP) that contain the replication parameters as its own...

That is just the nature of the copula algorithm. The transformation from continuous data (where the data are generated) to the binary outcome loses a lot of information, so correlation...

That is an improvement - and yes, your intuition is correct. I had a [post](https://www.rdatagen.net/post/simstudy-update-to-version-0-1-10/) about the method a while back, with a link to the [Emrich & Piedmonte](https://www.tandfonline.com/doi/abs/10.1080/00031305.1991.10475828) paper.