Hi @non87,
Thanks very much for your suggestions! I’m pleased to say both worked. It should have occurred to me that this parametrisation of sigma means that, for these choice of sigma values, sigma = 1.0, given that the treatment vector W values were only 0 or 1.
The other solution also worked really well; I didn’t realise that there was a start option when running pm.sample, but I’m glad I know now!
I’ll leave the thread open for a little longer to see if there’s any other ways to fix this other than just kick the starting values in the right direction, but thank you again!