BARTLETT ADJUSTMENT FOR SPARSE BINARY DATA

Mohammad Reza Zadkarami (Iran)

Received December 15, 2007

Abstract

In the present paper, we investigate the effect of sparseness of data on Bartlett adjustment. The choice of statistical tools for inference is highly dependent on the nature of the data. When the real data set is sparse, it creates some problems in using asymptotic theory. This property of the data restricts the use of statistical tools for inference, especially when the data set is extensive. The Bartlett adjustment factor improves the chi-square approximation to the distribution of the likelihood ratio statistic. We introduce Bartlett adjustment for binary response with probit link function. We find that the Bartlett adjustment does not improve the results when the binary data set is sparse.

Keywords and phrases: Bartlett adjustment, binary, likelihood ratio test, generalized linear models, sparse data.