Q4b. "Centurian Bank' has got a credit card division. They have used Discriminant Analysis to screen applicants so as to classify them as 'low credit risk' (group 1) or "high credit risk' (group2), based on the information on three variables collected from applications for credit card. Data on 1000 existing customers is available with the bank. The three variables considered to classify the customers are (1) Age, (2) Monthly Income and (3) Household Size. SPSS output for Discriminant Analysis is given below: Classification Results Wilks' Lambda Predicted Group Test of Wilks' Chi- df Sig. Membership Total Function(s) Lambda square 1 2 1 0.337 15.760 3 0.0013 Group Original 1 424 76 500 Count 2 44 456 500 2 % 1 2. 84.80 8.80 33 15.20 91.20 100.00 100.00 a 88.00 % of original grouped cases correctly classified. Standardized Canonical Discriminant Function Coefficients Function 1 Age -0.77738 Monthly Income -0.73963 Household Size -0.22719 Canonical Discriminant Function Coefficients Function 1 Age -0.09242 Monthly Income -0.00006 Household Size -0.13986 (Constant) 4.95839 Unstandardized Coefficients Tests of Equality of Group Means Functions at Group Centroids Wilks' Lambda F Age 0.621 59.900 Monthly Income 0.811 22.890 Household Size 0.837 19.075 df1 1 1 1 df2 998 998 998 Sig. .000 .000 .000 Function 1 1 1.321 2 -1.321 Unstandardized canonical discriminant functions evaluated at group means Give you response to the following on the basis of above output. (1) How good is the model? How many of the 1000 data points it classifies correctly? (ii) Which variables are relatively better in grouping the applicants as low risk/ high risk? (ii) If an applicant of credit card has age of 48 years, monthly income of Rs. 35,000, and household size of 5 family members, will you classify him as low credit risk or high credit risk? Why? (2.5 X 3 = 7.5) Q4b. "Centurian Bank' has got a credit card division. They have used Discriminant Analysis to screen applicants so as to classify them as 'low credit risk' (group 1) or "high credit risk' (group2), based on the information on three variables collected from applications for credit card. Data on 1000 existing customers is available with the bank. The three variables considered to classify the customers are (1) Age, (2) Monthly Income and (3) Household Size. SPSS output for Discriminant Analysis is given below: Classification Results Wilks' Lambda Predicted Group Test of Wilks' Chi- df Sig. Membership Total Function(s) Lambda square 1 2 1 0.337 15.760 3 0.0013 Group Original 1 424 76 500 Count 2 44 456 500 2 % 1 2. 84.80 8.80 33 15.20 91.20 100.00 100.00 a 88.00 % of original grouped cases correctly classified. Standardized Canonical Discriminant Function Coefficients Function 1 Age -0.77738 Monthly Income -0.73963 Household Size -0.22719 Canonical Discriminant Function Coefficients Function 1 Age -0.09242 Monthly Income -0.00006 Household Size -0.13986 (Constant) 4.95839 Unstandardized Coefficients Tests of Equality of Group Means Functions at Group Centroids Wilks' Lambda F Age 0.621 59.900 Monthly Income 0.811 22.890 Household Size 0.837 19.075 df1 1 1 1 df2 998 998 998 Sig. .000 .000 .000 Function 1 1 1.321 2 -1.321 Unstandardized canonical discriminant functions evaluated at group means Give you response to the following on the basis of above output. (1) How good is the model? How many of the 1000 data points it classifies correctly? (ii) Which variables are relatively better in grouping the applicants as low risk/ high risk? (ii) If an applicant of credit card has age of 48 years, monthly income of Rs. 35,000, and household size of 5 family members, will you classify him as low credit risk or high credit risk? Why? (2.5 X 3 = 7.5)
