Goodness of fit


Have the diagnostics of included in PyMC3? Or, are such diagnostics still recommended?


Many model fitting evaluation methods are available, Posterior Predictive Checks is a similar concept compare to goodness of fit in pymc2. Also see the explanation of WAIC and LOO in the doc

The new tools are super cool. Are suggesting that the diagnostics included with PyMC2 is unappealing or obsolete/invalid?


The diagnostics included in PyMC2 is probably quite limit and outdated - there are many new theories developed (like LOO) that are not available at the time.

OK, it seems that the Freeman-Tukey statistic has not been included. So I assume that is not recommened anymore…