This approach classifies the categorical dependent Y variable using characteristic X variables via Quadratic Discriminant Analysis (QDA) as shown in Figure 9.73. This method is similar to LDA, but the covariance matrix is used in the group assignment as well as the estimated coefficients because LDA assumes homoskedasticity in the prediction errors whereas QDA allows for some heteroskedasticity. This allows for second-order and second-moment approximations to calibrate the relevant group assignments. To get started, enter the variables you need to classify and enter the number of clusters desired. For instance, the required model inputs look like the following:
Figure 9.73: AI/ML Quadratic Discriminant Analysis
Notice that the forecasted classification groups for the QDA model below are identical to the LDA model shown previously. While the LDA model’s category is easily predicted using a multiple regression equation and looking selecting the category with the highest likelihood result, QDA requires the inclusion of the inverse covariance matrix. This means you will have to rely on the results presented and not readily be able to compute the likelihood results directly.
