Rolling Regression in PyMC

Thanks @jessegrabowski, this is super helpful. I ran the code and it works nicely. I’m trying to replicate the plot that you shared. I managed to plot a subset of the parameters using this code:

az.plot_forest(
    idata.posterior.stack(sample=("chain", "draw")).sel(regression=[0,1,2,3,4]),
    var_names=["alpha", "beta"]
)

However, I do not know how you added the parameters from the RollingOLS models and reduced the numbers of chains to a single one. Can you give me a hint on this?