For a long time I have been using a pretty neat trick to remove noise/outliers for skewed distribution, where I just access the whiskers from matplotlib’s pyplot and filter the data.