Through Monte Carlo Simulation, this study explores the approximate behavior of the two sample winsorized t test. Samples are drawn from the normal population and symetrically winsorized up to 20%. The two independent samples winsorized t test is then calculated on each sample using Monte Carlo methods using 1,000,000 iterations. The t values are then sorted from low to high and the critical values for both one and two tailed tests identified at the 95th, 97.5th, 99th and 99.5th percentiles. A table of critical values is then created, which represents the approximate distribution of the winsorized t statistic. |