Abstract:
|
The two-sample location-scale model permits pairs of standardized random variables to have a common distribution. Couched in terms of an equation, this can be expressed as Y=m(Z)+s(Z) e, where m is an unspecified location parameter vector, s is an unspecified scale parameter vector, Z is a binary covariate, and e is the observation error, the distribution of which is completely unspecified. Function-based hypothesis testing refers to formal tests that would help decide whether or not two samples may have come from any location-scale family of distributions. A comparison between two approaches, one based on empirical characteristic functions (ECFs) and another on plug-in empirical likelihood (PEL), will be the focus. The ECF test applies only for the uncensored case currently while the PEL test can be applied to both uncensored and censored data. An extension of the ECF test to the censored case will be discussed. Results of numerical studies will be reported.
|