.

# Comparison of statistical software

Interesting comparison table and comments, regarding the following statistical packages: R, MATLAB, SAS, STATA and SPSS. I wish Statistica would be included. The table tells you which statistical methods are available in each package. The list of statistical methods is itself impressive. Note that Jackknife (a resampling method in the table below) has nothing to do with Jackknife regression (new technique not implemented in any package yet, though Dr. Granville has promised to provide the source code).

Also statistical libraries are available in most programming languages, for instance Pandas in Python. Here are five interesting articles:

The table (below) and additional information about the various packages can be found here.

 TYPE OF STATISTICAL ANALYSIS R MATLAB SAS STATA SPSS Nonparametric Tests Yes Yes Yes Yes Yes T-test Yes Yes Yes Yes Yes ANOVA & MANOVA Yes Yes Yes Yes Yes ANCOVA & MANCOVA Yes Yes Yes Yes Yes Linear Regression Yes Yes Yes Yes Yes Generalized Least Squares Yes Yes Yes Yes Yes Ridge Regression Yes Yes Yes Lasso Yes Yes Yes Generalized Linear Models Yes Yes Yes Yes Yes Mixed Effects Models Yes Yes Yes Yes Yes Logistic Regression Yes Yes Yes Yes Yes Nonlinear Regression Yes Yes Yes Discriminant Analysis Yes Yes Yes Yes Yes Nearest Neighbor Yes Yes Yes Yes Factor & Principal Components Analysis Yes Yes Yes Yes Yes Copula Models Yes Yes Experimental Cross-Validation Yes Yes Yes Bayesian Statistics Yes Yes Limited Monte Carlo, Classic Methods Yes Yes Yes Yes Limited Markov Chain Monte Carlo Yes Yes Yes Bootstrap & Jackknife Yes Yes Yes Yes EM Algorithm Yes Yes Yes Missing Data Imputation Yes Yes Yes Yes Yes Outlier Diagnostics Yes Yes Yes Yes Yes Robust Estimation Yes Yes Yes Yes Longitudinal (Panel) Data Yes Yes Yes Yes Limited Survival Analysis Yes Yes Yes Yes Yes Path Analysis Yes Yes Yes Propensity Score Matching Yes Yes Limited Limited Stratified Samples (Survey Data) Yes Yes Yes Yes Yes Experimental Design Yes Yes Quality Control Yes Yes Yes Yes Reliability Theory Yes Yes Yes Yes Yes Univariate Time Series Yes Yes Yes Yes Limited Multivariate Time Series Yes Yes Yes Yes Markov Chains Yes Yes Hidden Markov Models Yes Yes Stochastic Volatility Models Yes Yes Limited Limited Limited Diffusions Yes Yes Counting Processes Yes Yes Yes Filtering Yes Yes Limited Limited Instrumental Variables Yes Yes Yes Yes Simultaneous Equations Yes Yes Yes Yes Splines Yes Yes Yes Yes Nonparametric Smoothing Methods Yes Yes Yes Yes Extreme Value Theory Yes Yes Variance Stabilization Yes Yes Cluster Analysis Yes Yes Yes Yes Yes Neural Networks Yes Yes Yes Limited Classification & Regression Trees Yes Yes Yes Limited Boosting Classification & Regression Trees Yes Yes Random Forests Yes Yes Support Vector Machines Yes Yes Yes Signal Processing Yes Yes Wavelet Analysis Yes Yes Yes ROC Curves Yes Yes Yes Yes Yes Optimization Yes Yes Yes Limited

Views: 17519

Comment

Join Data Science Central

Comment by Sergey on August 26, 2014 at 8:20pm

Dear Dr. Granville, the Stata / SPSS implementation of non-linear regression is quite inflexible. The SPSS implementation is especially bad, allowing one to type only simple, non-recursive functions in a pop-up window. If you got a project about implementing a non-linear regression for a complex functional form, you would use R, Matlab or a similar programming language. Following the general vibe of responses, I changed the “Non-linear Regression / SPSS” field to “Limited” to avoid potential misinterpretations of the table. However, the truth is: the SPSS implementation of non-linear regression is unsatisfactory for most industry-level research.

Lasso is available in SPSS only as part of categorical regression, which does not cover linear regression and generalized linear models. So in 90% of real-life situations lasso is not there… Regarding AMOS, it is not part of the standard SPSS license and IBM is charging extra money for it. But I can see how one can make an argument that buying SPSS and AMOS together is still cheaper than buying the standard SAS portfolio. So I have updated the web-site accordingly. Thank you for your comments.

On a different note, I have updated the table recently to include recent advances, like attempts of the SAS Institute to address Boosting and Random Forests…. Any further feedback is appreciated.

Comment by Jeremy Benson on August 23, 2014 at 6:28pm

R does everything and is free.  I have used Matlab, SAS and SPSS.  Matlab and SAS are very good, but the biggest problem is that I can't install a version on my home computer freely.  Also, if I want something beyond the license that my company has purchased, then I have to go through a process to build a business case to get that "package".  If I want that for R, I just go to CRAN and download it.  The costs for R have mainly been books to learn R.  However, there are so many free resources on R that you can learn to do it without buying anything.

Comment by Chandrasekhara S. ("C.S.") Ganti on August 18, 2014 at 2:37am

Hello Vince G.,

Yes, it a well taken response, since it was a comparison of various software as remarked,

Yes, it is tricky comparison-- I must admit. I am not  heavy IT / CS but use all software to my advantage and in the proper context and use it for a good application -- proven over a long haul -- On just another note that  old style statisticians are die-hard. Thanks to all FRS and ASA  and ISI  pioneers.. from Fischer to Box , From H. Cramer to Sir C.R/ Rao, from Shewhart to Deming, their contributions are invaluable -- Big Data or not .. withstanding.

speaking on Gartner -- I religious follow theirs and see latest comparison  for the BASEL compliance areas, I am looking forward to your take on that quadrant posting of various Software vendors.

**I just merely observed. In fact, I have not been in SAS for a bit.. Thx.

Comment by Vincent Granville on August 17, 2014 at 6:56pm

Hi Chandrasekhara - I've been using SAS for many years and it satisfied my needs. It does not matter the number of functions a package offers, only whether it offers what you are likely to use, and if it does it well. Here's an article on how to select a statistical package.

On a higher level, producing software reviews and comparisons is very tricky. Your reviews get outdated very quickly, and easily invite heavy criticism. Gartner sometimes provide interesting comparisons. I'm glad we have members here filling the gaps found in these reports.

Comment by Mark Samuel Tuttle on August 17, 2014 at 3:42pm

Mathematica should be included here.  Yes, I know it's proprietary, but one reason I pay for it (and use it) is the high degree of integration - a kind of almost everything is a "first-class object" notion.  This makes using the statistical stuff they have easier and more productive.  Regardless, I like seeing attempts at package evaluations.  A future evaluation would be include some notion of scalability.  I find myself doing ever bigger problems on my laptop; one reason is that my current laptop is much more powerful than past versions.

Comment by Chandrasekhara S. ("C.S.") Ganti on August 17, 2014 at 12:50pm

Please let us know if SAS does not have Experimental Designs  and Quality Control .. Is not .JMP a part of SAS ?? Please confirm .. Thanks

Comment by Chandrasekhara S. ("C.S.") Ganti on August 17, 2014 at 12:46pm

THE above table implies (???) SAS does not have non-parametric Tests. Please confirm if the YES is off the column to the right ??

Comment by Vincent Granville on August 15, 2014 at 10:10am

A reader said that this table has many errors about SPSS: Just starting at the top, contrary to what is indicated, SPSS has - and has had for years - ridge regression, lasso, nonlinear regression, path analysis (Amos) and more.