The fitdist module/script finds the probability distributions that best-fit a set of data. By default, eighteen common continuous distributions are examined for a best-fit. Setting dist=all results in over 90 continuous distributions being examined.
The Kolmogorov-Smirnov (KS) Goodness-of-Fit (GoF) test is used to determine whether a fit is suitable and to sort the results. If the KS GoF statistic is larger than the input (or default) threshold, fitpval, the distribution passes, otherwise it fails. For those distributions that pass, the following information is output: distribution name, list of MLE parameter estimates (in scipy.stat's defined order), and the KS GoF p-value.
A log file of distribution names, their MLE parameter estimates, the KS GoF statistic and p-value, and whether they pass or fail is always written out (see the logfile option).