Btw, if you have ESSs, then lots of other things can be improved. For example, the KDE that smooths samples into a continuous distribution curve (for tMRCAs and, I think, the other traces) somewhere uses the number of samples to decide the width of the smoothing Gaussian. That should really be the ESS (= number of independent samples), not the raw number of samples.
Btw, if you have ESSs, then lots of other things can be improved. For example, the KDE that smooths samples into a continuous distribution curve (for tMRCAs and, I think, the other traces) somewhere uses the number of samples to decide the width of the smoothing Gaussian. That should really be the ESS (= number of independent samples), not the raw number of samples.