New Statistical Procedures in TESTIMATE
Su-Wei test, Wei-Lachin procedure and Efron test in TESTIMATE. Also new exact binomial confidence interval and extended Cochran-Armitage test.
Su-Wei Test for difference or quotient of medians (both censored and not censored data are possible!): these tests are now available for testing equivalence and non-inferiority as well as difference testing. The median implemented here and the median test react to very small group differences even if the data contain many ties.
Wei-Lachin Procedure: idv previously provided this procedure as a standalone program ('SmarTest'). We have now integrated this program into TESTIMATE with no price increase. The Wei-Lachin procedure is particularly useful in connection with Phase-II Studies because the test is robust and the required sample size can be reduced by half when several variables are being used (Proof-of-Concept Study). Considered from a different point of view, however, the chance that the study is successful increases when the sample size remains constant. For example if the probability of success of a study is 80% (4:1), it will increase to more than 95% (19:1). The associated estimator together with the confidence region are included in the output, which makes it possible to compute a measure for the combination of efficacy and safety of a product (benefit/risk ratio).
Efron Test (1967) is now available for performing the Wilcoxon test with ties and censored data. The procedure provides the Mann-Whitney estimator and the average hazard ratio with the associated confidence intervals and also tests for non-inferiority and relevant superiority.
Binomial test: idv has replaced the procedure for the exact confidence interval associated with the binomial distribution in the 'One Group Procedure' module with a completely new procedure. Frequencies up to 100 000 can now be evaluated with a precision of 7 significant digits. Also, exact negative binomial and Poisson distributions are now available.
Cochran-Armitage Test: The weighing factors (arbitrary coefficients) in the Cochran-Armitage test can now also be processed weighted with group size. Thus, the now popular quotient test for non-inferiority with success frequencies in the so-called Gold Standard design with three groups can easily be calculated.
( see examples of applications).