Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.python > #57072 > unrolled thread

Re: Python package statistics

Back to article view | Back to comp.lang.python

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: Python package statistics Terry Reedy <tjreedy@udel.edu> - 2013-10-18 14:54 -0400

#57072 — Re: Python package statistics

On 10/18/2013 8:41 AM, Yaşar Arabacı wrote:
> Hi people,
>
> I collected some data on PyPI and published some statistics about
> packages on PyPI. I think you might find it an interesting read:
>
> http://ysar.net/python/python-package-statistics.html

"b2gpopulate (36MB)
...
Total sizes on packages in PyPI amounted to 4.2 GB. Average package size 
is 161 KB and standard deviation is 1MB."

For such highly skewed data, the mean and especially the standard 
deviation and confidence intervals are meaningless. The are 
'parameteric' statistics, which is to say, were designed for bell-shaped 
distributions. (I will not say 'normal' == Guassian distributions 
because they are *not* normal for much raw data.)

  A better summary is obtained from either 'non-parametric' statistics 
(median, inter-quartile range) or from 'normalizing' the data (if 
possible). For the latter, try taking the square root or log of the 
sizes and plot the distribution. If either works, take the mean and sd 
of the transformed values. Then report those and also the transformed 
back mean and mean+-sd.

-- 
Terry Jan Reedy

[toc] | [standalone]

Back to top | Article view | comp.lang.python

csiph-web