Pynnonen, S. (1992). Stepwise variable selection in quadratic discriminant analysis. Statistica Applicata 4, No. 4, 443--452.

Abstract
In this paper variable selection in a quadratic discriminant function between two normal populations is considered. Pynnönen (1987) has suggested the use of criteria like Akaike's AIC for selection of variables in discrimination between normal populations with unequal covariance matrices. In this paper to decrease the efforts of finding the final combination of variables in the quadratic discriminant function a stepwise procedure based on the above mentioned AIC-type criteria is suggested. Also some decision theoretic properties of the procedure are discussed. Finally for illustration purposes a numerical example is presented.