Reference no: EM133484314
Assignment: Marketing
Textbook: Building Better Models with JMP Pro (Chapter 5)
Use the Equity.jmp data from Blackboard for this exercise. This data set was first introduced. Recall that the response variable is the variable BAD, where the value 1 indicates that the customer is a bad credit risk.
1) Use functionality like the Columns Viewer, Distribution, Graph Builder, and Multivariate (Correlation) to re-familiarize yourself with this data.
Question I. Do any variables appear to be related to BAD? Explain in business language (it is not required to be technical in your response).
Question II. List any potential data quality issues you observe. There is no requirement to apply fixes. Only share your observations and possible recommendations.
2) Fit a logistic regression model for BAD, including all predictor variables.
Question I. What is the p-value for the whole model test?
Question II. What is the misclassification rate?
Question III. What are the two types of misclassification error that can occur in this example?
Question IV. How many misclassifications of each type were made?
Question V. Use the Effect Summary table to slowly remove non-significant terms from the model. How many terms are in your final model? Please include Effect Summary screenshot.
Question VI. What is the misclassification rate for this reduced model (include screenshot)?
Question VII. In the context of this example, define the two types of classification error: false positive and false negative. Which type of classification error occurred more often? What does this mean about our model? Explain.
Question VIII. What are the estimates (coefficients) for DEROG and CLAGE(include screenshot)? Ensure that you have applied Value Ordering to the dependent (target) variable, and the model is predicting the probability of an observation to be a "1", not a "0". In addition, open the Prediction Profiler, and explore what happens to the predicted probability that BAD=1 as you increase and decrease the values of these two variables.
Question IX. Continue to leverage the Prediction Profiler. What is the probability of being a BAD Risk individual when DEROG has a value of 2, and CLAGE has a value of 100 (leave all other variables at their original settings).
Question X. You need to explain what the coefficients for DEROG and CLAGE represent to your manager. Share your interpretation of the coefficients for these two variables (in non-technical terms).
Question XI. Save the Probability Formula for your model. Return back to your data table and observe the newly generated columns. Now focus on Row 20 in your table.
Question i. The current prediction for this individual is a Good or Bad risk?
Question ii. Controlling for all other predictors, begin to change the value for DEROG (number of derogatory reports) on Row 20 within the table itself. How many derogatory reports does this individual need before we would predict them to be a Bad risk?