Sankhya: The Indian Journal of Statistics

1986, Volume 48, Series A, Pt. 3, pp. 339--353

THE ASYMPTOTIC BEHAVIOR OF THE LIKELIHOOD RATIO STATISTIC FOR TESTING A SHIFT IN MEAN IN A SEQUENCE OF INDEPENDENT NORMAL VARIATES

By

YI-CHING YAO

And

  RICHARD A.  DAVIS, Colorado State University

SUMMARY. Let X1, …Xn be an independent sequence of random variables such that X1,… Xr ~ i.i.d N (m, s2) and Xr+1, …, Xn ~ i.i.d N(m + q, s2) where m, q ¹ 0, s2 and r are unknown parameters. The asymptotic properties of the likelihood ratio in testing H0 : r = n (no change point) vs. H1 : r < n are derived. It is shown, using a result of Darling and Erdos, that the likelihood ratio, suitably normalized and under H0 converges in distribution to the double exponential extreme value distribution. The asymptotic operating characteristics of the likelihood ratio test are studied and comparisons are made between the likelihood ratio test and a Bayes test. In particular, the Bayes test outperforms the likelihood ratio test as long as the change point does not occur very early or late in the sequence. Also, an approximation procedure to the null distribution is proposed which is not only correct in the limit and easy to compute, but provides a good approximation for 20 £ n £ 50.

AMS (1980) subject classification. 62F05, 62E20

Key words and phrases. Change point , Normalized Brownian bridge, Ornstein-Uhlenbeck process, Extreme value, Asymptotics

This article in Mathematical Reviews