A mortgage lender is concerned about rising default rates by homeowners. To minimise the risk of making 'bad' lending decisions in future, the mortgage lender decided to conduct a discriminant analysis using information on existing customers, some of whom had previously defaulted on their mortgage payments, coded with a default status of 1 (non-defaulters are coded with a default status of 0). The mortgage lender initially hypothesised that the following would be good predictor variables: ? age in years ? highest level of education (1 = school, 2 = undergraduate, 3 = postgraduate, 4 = postgraduate research). ? income in 000s ? outstanding credit card debt ? total years in employment. Analyse the selected SPSS output in Figure 3 (spread over the next two pages) and discuss what conclusions can be drawn from the data. In your analysis, be sure to address at least the following: ? Comment on the relative importance of the predictor variables. ? Comment on the suitability of including the 'highest level of education' variable. ? Determine the predictive accuracy of the model.

Figure 3 Eigenvalues Canonical Function Eigenvalue % of Variance Cumulative % Correlation 1 .217 100.0 100.0 422 Wilks' Lambda Wilks' Test of Function(s) Lambda Chi-square df Sig 822 97.203 5 000 Tests of Equality of Group Means Wilks' Lambda F df1 df2 Sig. Age .991 4.611 498 032 Years employed 937 33.289 498 000 Income .996 2.203 498 138 Credit card debt .957 22.489 498 000 Highest level of education 982 9.199 498 003 Standardized Canonical Discriminant Function CoefficientsStandardized Canonical Discriminant Function Coefficients Function 1 Age -.012 Years employed -.946 Income -.041 Credit card debt .930 Highest level of education .143 Structure Matrix Function 1 Years employed -.555 Credit card debt .456 Highest level of education 292 Age -.207 Income -.143Figure 3 (continued) Pooled Within-Groups Matrices Years Credit card Highest level Age employed Income debt of education Correlation Age 1.000 507 512 327 011 Years employed .507 1.000 669 477 -.140 Income .512 669 1.000 546 202 Credit card debt .327 .477 546 1.000 026 Highest level of education .011 -.140 .202 .026 1.000 Canonical Discriminant Function Coefficients Function 1 Age -.002 Years employed -.144 Income .001 Credit card debt 497 Highest level of education .165 (Constant) .411 Unstandardized coefficientsFunctions at Group Centroids Function Default status 1 0 -.195 1.106 Classification Resultsac Predicted Group Membership Default status 0 1 Total Original Count 0 315 110 425 16 59 75 9% 74.1 25.9 100.0 21.3 78.7 100.0 Cross-validatedb Count 0 312 113 425 19 56 75 9% 0 73.4 26.6 100.0 -+ 25.3 74.7 100.0