Use the Relay train data to develop a model to predict customer retention. You may use logistic regression to predict the variable "retained". For each of the regressions listed below, estimate the model coefficients in the train data and predict retention using the estimated model in the test data. You will use the coefficients obtained from the model estimated using the train data to do this. Name this predicted value "pretain". Calculate the hit rate. This can be calculated as % of matches between the value of pretain and retained in the test data.

1. Use esent, eclickrate, avgorder, ordfreq, paperless, refill, doorstep as independent variables to estimate the model using train data. Report the model coefficients. Predict retention, and calculate hit rate in the test data.

2. Use avgorder, ordfreq, paperless, refill, doorstep as independent variables to estimate the model using train data. Report the model coefficients. Predict retention, and calculate hit rate in the test data.

3. Use esent alone as independent variables to estimate the model using train data. Report the model coefficients. Predict retention, and calculate hit rate in the test data.

4. Create a dummy variable called weekend which is 1 if favday is Friday, Saturday or Sunday, and 0 otherwise. Use esent, eclickrate, avgorder, ordfreq, paperless, refill, doorstep, and weekend as independent variables to estimate the model using train data. Report the model coefficients, and predict retention, and calculate hit rate in the test data. Answer the following questions based on the 4 regression outputs above.

5. Why is esent a strong predictor of retention? Do you see any issues with using esent as a predictor for retention? Recommend transformations of esent that can overcome the issues of using esent as a predictor.

6. Does the sign of the coefficients for avgorder, ordfreq, and weekend make sense? What consumer behavior explanation can you provide for the sign of these coefficients?

7. What are your recommendations to Relay Foods Management for improving their customer retention?

Model parameters (Variable retained): Wald Odds Odds Source Value Pr>Chi? Odds ratio ratio ratio -3.323 0.227 0.020 1.255 1.239 1.270 Wald Upper -2.981 0.239 0.028 0.001 3.265 Chi? error Square 0.954 0.068 194.180 Chi? error Square 0.104 820.331 Chi? error Square 0.175 378.003