If you ran it with a uniform prior on b, say b ~ N (0, 100), I imagine you would get better mixing. I would like to understand intuitively why solution #2 works, but I don’t currently.
If you ran it with a uniform prior on b, say b ~ N (0, 100), I imagine you would get better mixing. I would like to understand intuitively why solution #2 works, but I don’t currently.