Critical Values for the Cusumsq Statistic in Medium and Large Sized Samples.

Data series containing more than 200 observations are common in financial economics. The usefulness of the cusumsq test in such medium-sized samples has been hampered by the lack of tabulated confidence bounds and by the inaccuracy of asymptotic approximations. In this paper, the authors extend Durbin's table to encompass all practical sample sizes. They have empirically calculated asymptotic limits, which prove to be extremely accurate for sample sizes greater than sixty. A convenient algorithm for calculating P-values, which is useful even in small samples, is also referred to. Copyright 1994 by Blackwell Publishing Ltd