[Major statistical software usable in epidemiology].

Abstract

Quality assurance of biostatistical data analysis is becoming mandatory. The aim of this review of statistical software was to guide uses who are faced with the increasing number of these tools. A list of softwares was obtained from statisticians and computer scientists, computerized databases, computer science lay press, and commercial documents. The softwares were described and their possibilities measured. The following functions were studied: functions considered as essential (descriptive analysis, Student's t, Mann-Whitney or Wilcoxon, ANOVA, Kruskall-Wallis, Pearson's chi 2, Fisher's exact test, Pearson's r), survival and multivariate analysis (logistic regression, proportional hazards modelling, times series). User-friendliness and quality of graphs were also studied. Among the 220 listed softwares, 26 could be considered as general softwares and were extensively analyzed. Statistical softwares may be classified into four categories: easy-to-use general public softwares with limited functions, such as EPI INFO; simple reference softwares with satisfactory user-friendliness and performing most of the statistical analyses, such as PCSM, STATISTICA or STATA; complex reference softwares intended for statisticians (EPILOG PLUS, SAS); specific cases such as EGRET performing only multivariate analysis or SUPERANOVA performing only analysis of variance. Such a review should be regularly updated.