Tail probability approximations in the first-order noncircular autoregression

Abstract

SUMMARY A saddlepoint expansion is derived for the cumulative distribution function of the least squares estimator of the autoregressive coefficient. The new method is shown to be superior both analytically and numerically to many existing approximations. Unlike other methods the new approximation performs uniformly well over the whole range of the distribution including the extreme tail, even when the sample size is as small as 10.