A Generalized Discriminant Rule When Training Population and Test Population Differ on Their Descriptive Parameters

Christophe Biernacki, Farid Beninel, Vincent Bretagnolle
2002 Biometrics  
Standard discriminant analysis methods make the assumption that both the labeled sample used to estimate the discriminant rule and the non-labeled sample on which this rule is applied arise from the same population. In this work, we consider the case where the two populations are slightly different. In the multinormal context, we establish that both populations are linked through linear mapping. Estimation of the non-labeled sample discriminant rule is then obtained by estimating parameters of
more » ... his linear relationship. Several models describing this relation are proposed, and associated estimated parameters are given. An experimental illustration is also provided, in which sex of birds which differ morphometrically over their geographical range is to be determined, and a comparison with the standard allocation rule is performed. Extension to a partially-labeled sample is also discussed.
doi:10.1111/j.0006-341x.2002.00387.x pmid:12071412 fatcat:56al6rmklrawblidzbet2illd4