Abstract
The distribution of the Quandt likelihood ratio λt. for a two-phase regression, has yet to be determined. In particular it is known that-2 log λt. is not distributed as chi-square (Quandt; 1960), for unknown switch point.
In this paper we describe sampling experiments which suggest that-log λt, has a Pearson Type III distribution, The parameters of the distribution appear to depend not only on the values of the x-vector (Feder; 1968) but also its dimension k.
Keywords: