A CHI-SQUARE STATISTIC WITH RANDOM CELL BOUNDARIES.
Abstract
Given a family F(x1 theta) of sufficiently smooth distributions in k-dimensional space depending on an m-dimensional parameter theta, estimate theta by an asymptotically efficient estimator theta prime. Define a chi-square statistic on random cells whose boundaries depend on theta prime. The asymptotic distribution of such statistics is obtained, generalizing work of A. R. Roy for k=1. The limiting distribution depends on theta in general, but is independent of theta for location and location-scale families. Moreover, the distribution approaches chi square (M-m-1) as the number of cells M approaches infinity. A table of upper critical points is given for several values of M for the case where F(x1 theta) is univariate normal. This should be useful in testing goodness of fit to the normal family. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 1969
- Accession Number
- AD0696072
Entities
People
- David S. Moore
Organizations
- Purdue University