This paper is concerned with the discrimination between two stationary AR((1))
plus noise processes in the time domain approach. The distribution of the discriminant
function, when the differences between two stationary processes is concentrated in their
variances rather than their means, leads to a linear combination of one degree of freedom chisquare
random variables. The coefficents are calculated numerically. An analytic method is
given by other investigations for ARMA processes, and an extension is given here for AR((1))
plus noise processes which have two errors rather than one. The weights and cumulants of the
discriminant function are compared using a numerical method