The problem with pm.trace_cov() is that it creates a covariance matrix of all parameters. In case I am sampling say one parameter with a different sampler, I won’t be able to use it.
Thus, I am creating my own covariance matrix using np.cov() of only the relevant parameters.
Which order do the parameters have to be in?
The order of the trace? (the order of the columns when saving it as text backend)
or the order that is displyed as an output to the console, e.g. NUTS: [kappa, s, alpha, gamma, sigmasq] ?