Is There Significance beyond the T-Test?

Abstract

Choosing the most appropriate statistical test may be routine for statisticians, but not for clinicians. The t-test, a parametric statistical test, may be used inappropriately. This commentary describes the assumptions of and alternatives to the t-test.

T-tests are used to compare two groups of data that are from a continuous scale and normally distributed. Determining if data are normally distributed can be difficult but selected methods can be useful. Two nonparametric tests, the Mann-Whitney U test and the Wilcoxon rank sum test, may be more appropriate in analyzing data that are neither continuous nor normally distributed. Statistical results may vary with the test chosen. Investigators are responsible for using the appropriate statistical tests. Statisticians and texts can be consulted. Pharmacy educational and training programs may need further emphasis in the area of statistics.