I’ve often wondered which stats package is the most popular around the world. As most stats packages are owned by private companies, there is effectively no publically available information to help answer the question. One indirect way to address the issue is to tabulate the number of hits each company’s website gets. Using information from google’s AdPlanner, Compete, and Alexa, I created a table that lists the “popularity” of each stats package’s website. If we can infer website popularity with software popularity, then this information may help answer the question.
As can be seen in the Table below, Mathematica, with 1,100,000 unique visitors per month, may be considered to be the most popular stats package, followed by Matlab and SAS. For whatever reason, data for SPSS is not available via AdPlanner (do they opt out?), so one has to estimate using Compete data, which is based on US visits only. As SPSS is a US company, I suspect ranking it just below Statisca is about right, although the Alexa World Rank suggests it should be just above Stata.
Table 1. Comprehensive Stats Packages and Corresponding Site Popularity Metics
Package
|
URL
|
AdPlanner
|
Compete
|
Alexa Rank
|
Mathematica
|
wolfram.com
|
1,100,000
|
309,073
|
7,494
|
SAS
|
sas.com
|
570,000
|
112,650
|
13,705
|
Matlab
|
mathworks.com
|
750,000
|
108,054
|
9,335
|
R
|
r-project.org
|
240,000
|
45,768
|
39,347
|
Statistica
|
statsoft.com
|
160,000
|
38,147
|
85,995
|
SPSS
|
spss.com
|
n/a
|
28,289
|
58,198
|
Minitab
|
minitab.com
|
85,000
|
22,172
|
146,970
|
Stata
|
stata.com
|
92,000
|
21,716
|
99,901
|
JMP
|
jmp.com
|
39,000
|
10,797
|
346,126
|
Systat
|
systat.com
|
Lack data
|
2,535
|
1,100,294
|
NCSS
|
ncss.com
|
Lack data
|
1,478
|
2,309,000
|
Notes: AdPlanner = Unique worldwide visitors per month; Compete = unique US visitors per month; Alexa Rank = based on a combination of average daily visitors (worldwide) to the site and page views on the site over the past 3 months; all data were obtained on May 20, 2011.
Rounding out the bottom are Systat and NCSS, which are very significantly less popular than the other packages. In some sense, I find this unfortunate, as both of these packages are not that much less useful or useable than their competitors. In some respects, they are on par or even better.
It will be interesting to keep an eye on R, the only freeware package included in the analysis. Will it one day capture the majority of the stats package market? Years ago, I thought it would, but now I seriously doubt it. I suppose that could be a topic for a future blog entry.
≈ Gilles