Class of Tests Based on Subsample Quantiles for Two-sample Scale Problem
Shashank D. Shindhe
Department of Statistics, Karnatak University, Dharwad, India.
Sharada V. Bhat *
Department of Statistics, Karnatak University, Dharwad, India.
Parameshwar V. Pandit
Department of Statistics, Bangalore University, Bangalore, India.
*Author to whom correspondence should be addressed.
Abstract
A class of distribution-free tests based on U-statistics with its kernel being function of subsample quantiles is proposed for a two-sample scale problem. The proposed class of tests is a general class of tests that includes numerous members which explore information from the tails of the distributions and tests resistant to outliers. This class of tests includes many existing classes of tests as its subclasses. The distribution of the proposed class of tests is derived and its relevance is discussed. One of its members, which is resistant to outliers in -sample is investigated in detail.
Keywords: Two-sample, quantiles, asymptotic relative efficiency (ARE), null distribution, empirical power