Comparison and Empirical Evaluation of Classical Tests of Skewness versus Bootstrap Tests of Skewness

This study compares classical (SK-1, SK-2, SSS, MED, STDM, PRSN)and bootstrap (KS and Student’s t) tests on the basis of size and power properties for six different data generating processes (chi-square distribution, beta distribution, lognormal distribution, mixture of two normal distribution, and mixture of two uniform and normal distribution) via Monte Carlo Simulations. In general, the classical tests for skewness perform better than bootstrap tests, however, in certain situations the bootstrap tests perform better. Therefore, this study recommends a strategy for choice of test to be applied in different situations. If the data histogram shows deviation from symmetry and the third moment is close to zero then the bootstrap tests should be used. In other cases the classical tests of skewness, in particular SK-2 which is the best performing test, should be used.