Skip to main content

Advertisement

Table 5 Estimated probabilities (%) that the p-value of the WMW test is smaller than that of the t-test, averaged over 32 combinations of amount of skewness and standard deviation ratios

From: t-tests, non-parametric tests, and large studies—a paradox of statistical practice?

  Number of subjects in each group
  10 25 50 100 250 500 1000
Gamma distributions 54.1 63.3 69.8 75.8 82.6 86.8 90.3
Lognormal distributions 45.6 52.1 56.9 62.3 70.1 76.3 82.5