Free statistical software

[terzi]

Hi all,

I am a pro free-soft person (although I don't think that adjective exists) so I've tried a few open source and free alternatives to perform statistical analysis. Although most people will think in R when they hear free statistical software (and all the programming involved with it), there are many great options that are very easy to use and that can can be very helpful.

Here are some of my personal choices.

OPENSTAT

This free alternative is designed to aid stats students who work in social sciences. It has a really easy to use menu based interface and it has many analysis available: regression, hypothesis testing (t-test, z-test, etc.) non-parametric analysis, multivariate regression, ANOVA.

VISTA

This package is mostly designed to create graphs that are not available in some other software. Although it may be a little tricky to use, once you learn you will have tons of graphing and analytical tools available for free.

KYPLOT

My personal favorite. I absolutely love this one. It is very similar to excel, you can even customize cells and add formulas in the same way. It has several statistical and mathematical tools and even some module to perform matrix operations. Time series analysis, non-parametric statistics, multivariate analysis and even survival analysis are included here. Still, there's nothing perfect. The newest versions are no longer free, but you can find the old ones that are still available for uncharged use.

MACANOVA

Although I didn't use this one much, it is designed mostly for ANOVA and MANOVA models. Recently it has been updated with more tools for multivariate analysis. It is not so simple to import data but after that the use isn't complicated.

POP TOOLS

An excel add-in that includes matrix operations, tools for simulation (monte carlo and I think some bootstrapping) and some statistical procedures.

RLPLOT

This is a light easy-to-use soft that can also run in Linux. It is mainly focused in graphing but it has some options to perform ANOVA and regression models. Graphs can be easily modified to obtain a professional look.

You can check these out and many more in the following link:

http://talkstats.com/newthread.php?do=newthread&f=8

It is true that any commercial software will overcome free software but for non-professional users or for students these are certainly a great alternative. I personally combine free soft with commercial ones at work.

Hope some of you find this useful.

[WeeG]

speaking of free software, did anyone try Mystat, the smaller version of Systat ?

[gianmarco]

Hi,

I would add PAST.

It is has been developed with the needs of palaeontologist in mind, but it performs many statistical analyses that are useful for many purposes.

Moreover, it is free.

Pasting from the home-page (http://folk.uio.no/ohammer/past/):

# A spreadsheet-type data entry form
# Both interactive user interface and scripting
# Graph, scatter, 3D scatter, bubble, histogram, kernel density estimation, box, percentile, ternary, survivorship, spindle, matrix, surface and normal probability plots
# Curve fitting: Linear (Standard and Reduced Major Axis) with bootstrapping and permutation, lin-log (exponential), log-log (allometric), polynomial, logistic, von Bertalanffy, sum-of-sines, smoothing splines, LOESS smoothing, Gaussian (species packing), multiple regression.
# F, t, permutation t, Chi-squared w. permutation test, Fisher's exact, Kolmogorov-Smirnov, Mann-Whitney, Shapiro-Wilk, Jarque-Bera, Spearman's Rho and Kendall's Tau tests with permutation, correlation, covariance, contingency tables, one-way and two-way ANOVA, one-way ANCOVA, Kruskal-Wallis test, sign test, Wilcoxon signed rank test with permutation, Fligner-Killeen test for coefficients of variation, mixture analysis, survival analysis (Kaplan-Meier curves, logrank and other tests), risk difference/risk ratio/odds ratio with tests.
# Diversity indices with bootstrapping and permutation, individual- and sample-based rarefaction. Capture-recapture richness estimators. Renyi diversity profiles, SHE analysis, beta diversity.
# Abundance model fitting: Geometric, log-series, log-normal, broken stick.
# Multivariate statistics: Principal Components (with Minimal Spanning Tree, bootstrapping etc.), Principal Coordinates (19 distance measures), Non-metric Multidimensional Scaling (19 distance measures), Detrended Correspondence Analysis, Canonical Correspondence Analysis, Cluster analysis (UPGMA, single linkage, Ward's method and neighbour joining, 19 distance measures, two-way clustering, bootstrapping), k-means clustering, seriation, discriminant analysis, one-way MANOVA, one-way and two-way ANOSIM, one-way NPMANOVA, Hotelling's T2, paired Hotelling's T2, Mahalanobis-distance permutation, Mardia's multivariate normality, Box's M, Canonical Variates Analysis, multivariate allometry with bootstrapping, Mantel test, SIMPER, Imbrie & Kipp factor analysis, Modern Analog Technique, two-block Partial Least Squares.
# Time series analysis: Spectral analysis, autocorrelation, cross-correlation, wavelet transform, Walsh transform, runs test. Mantel correlogram and periodogram. ARMA, Box-Jenkins intervention analysis. Solar forcing model.
# Geometrical analysis: Directional statistics (Rayleigh, Rao, chi-squared, Watson-Williams, circular kernel density estimation, angular mean with CI, rose plots, circular correlation), kernel density estimation of point density, point distribution statistics (nearest neighbour and Ripley's K), Fourier shape analysis, elliptic Fourier shape analysis, eigenshapes, landmark analysis with Bookstein and Procrustes fitting (2D and 3D), thin-plate spline transformation grids with expansions and principal strains, partial warps and scores, relative warps and scores, centroid size from landmarks, size removal by Burnaby's method.
# Parsimony analysis (cladistics): Exhaustive, branch-and-bound and heuristic algorithms, Wagner, Fitch and Dollo characters. Bootstrap, strict and majority rule consensus trees. Consistency and retention indices. Three stratigraphic congruency indices with permutation tests. Cladograms and phylograms.
# Biostratigraphy with the methods of Unitary Associations, Ranking-Scaling (RASC), Appearence Event Ordination and Constrained Optimization (CONOP). Confidence intervals on stratigraphic ranges.
# Gridding (spatial interpolation): Moving average, thin-plate spline and kriging with three semivariogram models.

Included in the distribution are real data sets for educational use, together with extensive documentation and case studies.

PAST has been tested under Windows 95, 98, 2000, NT 4, XP and Vista.



Regards,
Gm

[Dason]

Hi all,

I am a pro free-soft person (although I don't think that adjective exists) so I've tried a few open source and free alternatives to perform statistical analysis. Although most people will think in R when they hear free statistical software (and all the programming involved with it), there are many great options that are very easy to use and that can can be very helpful.

It's easy to overlook but there are quite a few GUI frontends to R that provide some menu based usability for those less inclined to learn the command line. Two that immediately pop into mind are

RKWard and R Commander.

I'm sure there are a few more (I actually know a grad student here at ISU that is creating a web based R interface but it's only in it's beta stage and not really funcitonal yet but it's still pretty cool). I'm an R person myself so that's just my little take on making R a little bit easier to the new user. Of course the ones I mentioned are free and I know RKWard is open source and I think R Commander is as well but I'm not positive so if you're into that sort of thing there's another plus.