Karin Gembert

SAS macros for computations of non parametric confidence intervals

Abstract

The testing of new drugs and aids can be made with different
statistical methods and with the help of statistical programs, one of
which is SAS. The data collected from clinical trials are often said
to be non parametric which requires special methods. Because of the
need for good computers these methods are not yet fully developed in
computer programs.

The aim of this paper is to make two programs, so-called macros, that
within SAS calculate non parametric intervals for both one and two
sample analyses. The one sample interval is based on the Wilcoxon
signed rank test and can, for example, be used to test a null
hypothesis claiming no change due to the drug or the aid against any
suitable alternative hypothesis. The two sample interval is based on
the Wilcoxon rank sum test and is thereby used to test if there is a
difference when comparing the new drug or aid with one already in
use. The null hypothesis for this will be that the population with the
new drug or aid has the same mean or median as the one tested against.