Over the past 15 years Cytel developed algorithms for exact statistics.
What are exact tests?
Exact tests are hypothesis tests that are guaranteed to preserve the Type I error rate of the test. In other words, these tests, when applied to a null model, will not produce "significant" results more than 5% of the time, if your test is conducted at the .05 level of significance. Permutation procedures in which the data are repeatedly rearranged (permuted) and the test statistic recalculated lie at the heart of most exact procedures, including those in StatXact and LogXact. Unlike standard procedures, which rely on approximations to the sampling distribution of a statistic, permutation procedures produce exact sampling distributions.
The products distributed by Timberlake Consultants are
- StatXact® and StatXact PROCs for SAS® Users
The most widely used software for small-sample categorical and nonparametric data. StatXact PROCs are fully integrated with SAS software.
- LogXact® and LogXact PROCs for SAS® Users
A breakthrough for small-sample logistic regression. LogXact PROCs are fully integrated with SAS software.
- Egret®
Intuitive tools for design and analysis of epidemiological studies
- ToxTools®
Software for Dose-Response Modeling, Benchmark Dose Estimation and Risk Assessment.