Sisvar

Description

Sisvar is a statistical analysis and design of experiments program. Its first version was launched in 1996 for the DOS operating system. The original language of its programming was Pascal and the compiler was the Borland Turbo Pascal 5.5.

Later, with the advent of Delphi, the Sisvar was converted to Windows. Its first Windows version was compiled in Delphi 1.0, using object-oriented Pascal language. Since then the Sisvar is being constantly updated and is currently compiled by Delphi 7.0. I am working in a Java Version of Sisvar. In this case, the software will present a english menu and report. Waite please.

Some basic characteristics of Sisvar:

  • Version: 5.6
  • Build: 86
  • Programming Language: object-oriented Pascal
  • Language: portuguese (Java version is coming in english)
  • Financial Support: CNPq (Researcher grants)
  • Operational System: Windows, even though it runs in Linux by using Wine to install (Windows Emulator)
  • Register: 28/04/2006 under the register number: 828459851
  • Sisvar is Free

Basically, Sisvar is capable to perform the following statistical analyses:

  • Analysis of variance (anova): the user can specify your own model of analysis of variance according to his/her needs. The software lets the user specify the model with up to 200 effects and 150 errors (350 sources of variation in the total). Some of these effects (sources of variation) can be intermediate errors such as in split plot and split-block designs. The Sisvar performs multiple comparison procedures for treatments means, regression analysis for treatments with quantitative levels, contrasts for structered effects, and much more. The capability of unfold the effect of interaction is the great advantage of Sisvar.
  • Experimental Designs: Sisvar allows the user to randomize his/her experiments and generates these plans for a completely randomized, randomized blocks and partially balanced incomplete blocks (PBIB) designs. In addition, the user can choose different treatments structures such as factor structures and split plot.
  • Kernel density estimators.
  • Descriptive Statistics: Sisvar allows user to obtain central tendency and dispersion measurements and descriptive statistics of the distribution nature. It also provides histograms and Shapiro-Wilk normality test.
  • Normality test of Shapiro-Wilk.
  • Probability compuations: computes densities, quantiles, cumulative probabilities for several probability models, including noncetral distributions.
  • Estimations and hypothesis tests: Sisvar performs point and interval estimation to several parameters of interest from one or two populations, independent or paired. For these same cases, also performs hypothesis tests.
  • Simple and multiple linear regression: Sisvar performs regression in the context of analysis of variance, using the pure error to calculate standard errors and to apply hypothesis tests. Similarly, it allows the user to make a regression study outside the experimental design context, i.e., it in a context not related with regression analysis of some source of variation within a experiment with replications (without being within a replicated experiment). The user can set different effects as simple mathematical functions. Moreover, Sisvar allows the selections of best linear models using the backward (my preferred), forward and stepwise techniques.
  • Bootstrap multiple comparison procedures: this is a powerful technique that Sisvar provides to its users. In this case, Sisvar can perform Dunnett's, Tukey and SNK test by means of bootstrap. Furthermore, Sisvar has also a plug-in method of the bootstrap Scott and Knott test. All these topics were the subject of my post-doctoral studies under the supervision of Professor Clarice Garcia Borges Demetrius and Professor Bryan F.J. Manly in the company of my friend and colleague Amauri de Almeida Machado, author of one of the most important statistical analysis software that has already being used in Brazil, the Sanest.

 

 

Download

  • The last version of Sisvar can be found here: Download.
  • Download Statistics: See it
  • IMPORTANT: Preferably, install Sisvar as an administrator by clicking the right of the setup file and choosing run as administrator.
  • Note: the above download link has a first part the is just a way do count the download numbers, redirecting it, first to the page of http://dstats.net/.

  • A little portugues manual of Sisvar: download
  • Intructions for save dbase (extension dbf) file in Excel 2003 ou early in portugues: download
  • Currently I strongly recommend the use of LibreOffice para to do that conversion. The files in dbf can be prompt used in Sisvar. It is easily performed in the LibreOffice. Just put the names of each column in the first row and the variables registers (values) in the other rows. Each column should be a file variable. the save the file as dbse and you are ready to go. download the LibreOffice.
  • My old yellow page nootebook with the combined variances in experimental designs with more than an error term. It should be used to unfold (slice) interaction effects into one of its effects. This is absolutely a non formal results. I am working in a presentable version with rigorous proofs: download it here.

Citations

Cite Sisvar by:

  • FERREIRA, Daniel Furtado. Sisvar: a Guide for its Bootstrap procedures in multiple comparisons. Ciênc. agrotec. [online]. 2014, vol.38, n.2 [citado  2015-10-17], pp. 109-112 . Disponible en: ISSN 1413-7054.  http://dx.doi.org/10.1590/S1413-70542014000200001.
  • Ferreira, Daniel Furtado. Sisvar: a computer statistical analysis system. Ciência e Agrotecnologia (UFLA), v. 35, n.6, p. 1039-1042, 2011.
  • link: article Ciência and Agrotecnologia
  • Ferreira, Daniel Furtado. SISVAR: um programa para análises e ensino de estatística. Revista Symposium (Lavras), v. 6, p. 36-41, 2008.
  • link: article Symposium
  • FERREIRA, D.F. Análises estatísticas por meio do Sisvar para Windows versão 4.0. In...45a Reunião Anual da Região Brasileira da Sociedade internacional de Biometria. UFSCar, São Carlos, SP, Julho de 2000. p.255-258.