Thanks @mike-lawrence - appreciate the response.
I’ve tried out the code you suggested. The model still doesn’t sample well (low ESS, high Rhats). Looks like each chain is still getting stuck.
I’ve also tried out fixing each parameter of the model apart from the random walk coefficients to see if they were contributing to any issues with non-identifiability etc., but it looks the random walk prior is causing the problems here.
My understanding from another thread is that the parameters will cause a dependence on n, even when there is no trend present.
I’m not sure if that is correct, but I am going to try this model on a simulated dataset with no trend in the coefficients and see what happens.