Testing the equality of a large number of U-estimable parameters.

M. Romero Madroñal, M. R. Sillero Denamiel, M. D. Jiménez Gamero

The problem of comparing a parameter, estimable through U-statistics, across k populations is addressed. The kernel is assumed to have degree two. A statistic is proposed, shown to be distribution-free under the null hypothesis and certain alternatives for large k. This property under the null allows for the construction of a test, while the distribution under alternatives enables a power analysis. The finite performance of the test is numerically assessed through a simulation study.

Keywords: Testing degree-2 U-statistics consistency

S. Saavedra Martínez, A. López Cheda, M. A. Jácome Pumar

X. de Juan Soriano, S. Mazuelas

D. Rodríguez Vítores, E. del Barrio Tellado, J. Loubes

M. J. Campuzano, J. J. Muñoz, V. Deibe Blanco