Comments on: Histograms — matplotlib vs. R http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/ cognition, language, social systems; statistics, visualization, computation Tue, 25 Nov 2025 13:11:20 +0000 hourly 1 By: brendano http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-171505 brendano Mon, 06 Aug 2012 14:52:07 +0000 http://brenocon.com/blog/?p=1112#comment-171505 O.R.: Good question… QQ Plots are good for continuous data, of course. You can use them for discrete data, if you impose and ordering on the levels/values. This does seem artificial though.

]]> By: O.R. http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-167869 O.R. Wed, 25 Jul 2012 14:27:40 +0000 http://brenocon.com/blog/?p=1112#comment-167869 I’m curious, if histograms aren’t optimal, what kind of exploratory visualization would you use to check the distribution shape of discrete data?

]]> By: Brendan O'Connor http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-120218 Brendan O'Connor Wed, 08 Feb 2012 19:05:16 +0000 http://brenocon.com/blog/?p=1112#comment-120218 Lukas — yeah it’s dumb to use histograms-designed-for-continuous-data at all; I just think the R histogram is doing better in a challenging situation.

]]> By: Lukas Biewald http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-120212 Lukas Biewald Wed, 08 Feb 2012 18:19:27 +0000 http://brenocon.com/blog/?p=1112#comment-120212 Brendan – are you sure the R plot is really better? Just looking at the first comparison you have – R magically picked a perfect bin size that doesn’t have aliasing effects with you underlying discrete data. But it completely hides the fact that your data is discrete. It sure looks prettier, but I think the prettiness is misleading – I would guess from that histogram that the data comes from a continuous distribution.

If the possible values of your data is a small set of discrete points, why use a histogram at all?

Anyway, just a thought :). Come back to California!

]]> By: brendano http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-118993 brendano Sat, 04 Feb 2012 05:19:08 +0000 http://brenocon.com/blog/?p=1112#comment-118993 OK I figured out my problem. RPy2 crashes like crazy. But RPy works nicely. (I installed rpy with https://gist.github.com/1732879 ). To get plotting to work, I have to start Python with “ipython –pylab”. And this seems to work for R’s graphics commands, not just matplotlib (which it’s supposed to be for). If you don’t say “–pylab”, then the plot command is unreliable.

]]> By: michael toomim http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-118703 michael toomim Fri, 03 Feb 2012 01:25:55 +0000 http://brenocon.com/blog/?p=1112#comment-118703 I used rpy. Rpy2 seemed lame.

]]> By: brendano http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-118683 brendano Thu, 02 Feb 2012 22:15:40 +0000 http://brenocon.com/blog/?p=1112#comment-118683 RPy crashes like crazy for me. Do you use “rpy” or “rpy2″ ?

]]> By: Yang http://brenocon.com/blog/2012/02/histograms-matplotlib-vs-r/#comment-118681 Yang Thu, 02 Feb 2012 22:05:35 +0000 http://brenocon.com/blog/?p=1112#comment-118681 You probably know about it already, but I like using RPy to leverage R for plotting and other “better-in-R” routines.

]]>