SIMSTAT: Analysis of proportions


This is chosen when you have y successes in N trials in several groups and you want to compare the alternative estimates for the binomial parameter, with exact confidence limits, by statistical and graphical methods. Consult program BINOMIAL for more details about the binomial distribution.
You can input a matrix with columns of y and N (as in the test file binomial.tf2) or you can add a third column for another variable (x, if the spacing is parametric as in the test file binomial.tf3). Note that exact error bars are calculated, where you can alter the default significance level, and data transformed into log odds for plotting has exact (unsymmetrical) transformed error bars.
If you are interested in comparing proportions across a set you can request tables for all the differences, ratios (relative risks), and odds ratios with large sample confidence limits.
In the special case where the data are a 2 by 2 by k set, that is k sets of adjacent 2 by 2 contingency tables, a Cochran-Mantel-Haenszel Meta Analysis can be done and the k Log Odds Ratios plotted.

Back to Help Menu or End Help