Currently there seems to be a difference with our implementation of normalising constant of the log prob in the lkj_correlation distribution. So, we need to find a different reference implementation of the normalising constant. This does not impact MCMC or sampling, but may affect future uses of lkj_correlation. For example, in the integration and marginalisation interface.
Currently there seems to be a difference with our implementation of normalising constant of the log prob in the lkj_correlation distribution. So, we need to find a different reference implementation of the normalising constant. This does not impact MCMC or sampling, but may affect future uses of lkj_correlation. For example, in the integration and marginalisation interface.