Conclusions so far

[maxbiostat]

• The hybrid implementation, which switches to the approximation of Gaunt et al. (2019) when mu > 1.5 and nu > 1 (slightly different criterion than the original implementation) seems to have trouble converging, whereas the purely adaptive implementation works fine.
• The brms implementation is roughly twice as fast as the adaptive one for mu=5 and nu=2. But for mu=5 and nu=0.5 the time difference comes down. In terms of ESS, adaptive is better for the fist configuration, but things are pretty much the same of the second.