Reduced-Rank Regression Models With Two Sets of Regressors
In Chapter 2 we have demonstrated the utility of the reduced-rank model for analyzing data on a large number of variables. The interrelationship between the dependent and independent variables can be explained parsimoniously through the assumption of a lower rank for the regression coefficient matrix. The basic model that was described in Chapter 2 assumes that the predictor variables are all grouped into one set and therefore, they are all subject to the same canonical calculations. In this chapter we broaden the scope of the reduced-rank model by entertaining the possibility that the predictor variables can be divided into two distinct sets with separate reduced-rank structures. Such an extension will be shown to have some interesting applications.
KeywordsLinear Discriminant Analysis Error Covariance Matrix Alternative Estimator Linear Discriminant Function Asymptotic Covariance Matrix
Unable to display preview. Download preview PDF.