Assignment:
Practice Exam No notes, calculators, or R programming will be allowed during this exam. No use of R is required to complete the questions below.
SCENARIO 1
You are the CHRO of Kramerica Industries, a consulting firm. You are tasked with increasing employee productivity AND improving hiring practices over the next eighteen months. Use the dataset described below to answer the questions and develop a plan of action for each. The appendix has all of the information you'll need to answer each question.
Variable Description technical 1 indicates this employee has a technical background, 0 otherwise (0 could be an HR role, an administrative role, etc.) yearsofservice number of years the employee has worked for the firm (0 indicates a new, entry-level employee) currentsalary total annual salary for each employee at the firm performancereview values of 1-10 with 10 being an excellent review at the end of last year leadershiplevel values of 1-5 where 5 is the highest level of promotion and 1 is entry level levelofeducation values of 1-5 where 5 is PhD or similar, 4 is MS, MBA or similar, 3 is college graduate, 2 is some college, and 1 is high school graduate certifications number of professional certifications held by employee peerreviews values of 1-10 with 10 being an excellent peer review at the end of last year.
1. You ask your analysts to run a model of employee performance to better understand employee motivations. The model is based on whether the employee has a technical role, years of service, current salary, education, certifications, and peer relationships. Use Appendix 1 for all questions.
a. Which variables are significant predictors of performance?
b. Interpret the coefficient of the technical role indicator. Does this coefficient make sense? Under the assumption we have a good linear model, should we remove the technical indicator? Use your intuition and the group means to answer this question. Answer the remaining questions based on the correct model (with or without the technical indicator).
c. Interpret the intercept. Answer the remaining questions based on the correct model (with or without an intercept).
d. Interpret the coefficient of current salary. Is it economically significant? (Hint: Statistical significance and economical significance are different! Economical significance is "real world" significance.)
e. Is multicollinearity an issue here? How do you know? Which coefficients can be interpreted directly and which require more careful evaluation?
f. Test each of the five model assumptions and describe how you evaluated each
i. Include specific references to charts and other analytical results.
ii. Be sure to rely on intuition!
iii. If any assumptions are violated, explain what issues may exist and how they can be addressed.
iv. Note: Charts used for testing assumptions are created based on a single model, though you have chosen one of several in the above questions.
g. Which variables seem to increase performance the most?
h. How much of the variation in performance is left unexplained? Are there other variables that might predict performance that were not included in the model?
i. You have a budget of $100,000 to spend on employee enrichment with the goal of improving overall productivity. With it, you plan to launch an internal campaign for improved performance of employees. What is at the center of your campaign?
2. With the goal of improving hiring practices, you want to predict performance of future employees from things you can know ahead of time (previous salary, peer reviews from references, resume details, etc.). You ask your analysts to generate a predictive model similar to the one above. Use Appendix 1 for all questions.
a. How do you suggest they divide the data into test and training sets?
b. Is this model a candidate for a variable selection method? Why or why not?
c. Do we need to worry about outliers in performance here? Why or why not?
d. Would you use this predictive model if it had an average error of 1 point (in a performance review) in the test set? Why or why not? What about an average error of 3 points?
SCENARIO 2
You are the Operations Manager of FedEx distribution centers in the US. In an effort to improve daily delivery efficiency, you've asked the Operations Analytics team to create a couple of models for you. The models are included in the appendices. The data used is described below.
Variable Description driversworking total number of drivers employed by this firm who are delivering packages on this date weekend 1 indicates this observation is on a weekend, 0 otherwise expectedpackagesdelivered total number of packages planned for delivery on this date extrahands 1 if an additional 1,000 workers should have been hired temporarily for this day weatherconditions 100% indicates perfect weather, 0% indicates bad weather (snow, no packages delivered) pctoversized percent of packages that are oversized on this date 1. The model shown in Appendix 2 includes two predictor variables and their interaction.
a. Why are we using a linear regression here?
b. Which variables are significant predictors of driversworking?
c. Write out the two equations associated with this model. (Hint: What happens when weekend = 0 versus weekend = 1. This is only possible because weekend is binary, though other types of interactions are possible.)
2. A new model is shown in Appendix 3.
a. Why are we using a logistic regression here?
b. Which variables are significant predictors of the log odds ratio of extrahands and which coefficients should be assigned to 0?
3. Write out the equation associated with the model in Appendix 4 and interpret each coefficient's effect on the log odds ratio. (Same model as Appendix 3, but only significant predictors are included.)
4. Using the model shown in Appendix 4, write out the steps and equation(s) you might use to make a prediction concerning whether you need extra workers today. Today, there are 11,543,100 packages expecting to be delivered, weather is at about 65% due to heavy rain, and about 3% of the packages are oversized.