What's so hard about histograms?

Re:What's so hard about histograms? by svara on Saturday July 22, 2017 @11:15AM (#14827584)

When you think you want to plot a histogram, it's often a better idea to plot a (empirical) cumulative distribution [0] instead. You don't have to worry about how to select your bin limits and you can usually put several in the same plot for comparison without making it unreadable due to overlap.

[0] https://en.wikipedia.org/wiki/Empirical_distribution_functio....

[ Parent | on HN ]

Re:What's so hard about histograms? by lukego on Saturday July 22, 2017 @12:04PM (#14827823)

I'm also enarmored of ECDFs lately. I was really happy to discover the KS-test for making comparisons between them.

https://en.m.wikipedia.org/wiki/Kolmogorov–Smirnov_test

https://stats.stackexchange.com/questions/288416/non-paramet...

[ Parent | on HN ]

Re:What's so hard about histograms? by aji on Saturday July 22, 2017 @11:19AM (#14827609)

I like using cumulative distributions because they make small changes in the data a little more obvious. e.g. if all the buckets are 10 but there's a section where they're 11, that difference will show up in a cumulative distribution as a bend in an otherwise straight line, which in my opinion is a much easier difference to see

[ Parent | on HN ]

Re:What's so hard about histograms? by convolvatron on Saturday July 22, 2017 @03:21PM (#14828851)

isn't seeing a little hump in a pdf clearer?

[ Parent | on HN ]