Local asympotic efficiency of some goodness-of-fit tests under skew alternatives.

The efficiency of most known distribution-free goodness-of-fitt tests such as Kolmogorov - Smirnov, Cramér - von Mises and their numerous variants was studied mainly under the classical alternatives of location and scale. However it is interesting to compare the efficiencies of these tests under asymmetric alternatives among which the most popular is the so-called skew alternative proposed by Azzalini (1985) in the case of the normal distribution. We find and compare local Bahadur efficiencies of many known statistics for skew alternatives and discuss also the conditions of their local optimality.