1- The sales department is interested in knowing the factors (e.g. employee tenure, employee salary) related to the amount of sales (in dollars) each salesperson generates. Given the outcome variable "Number of new clients", which of the following regression techniques is most likely appropriate?

a. Poisson Regression

b. Linear Regression

c. Logistic Regression

2- A logistic model was built to predict whether or not students are likely to graduate (1 = graduate, 0 = not graduate). When examining the lift chart, at 10% of the population the y value was 1.3. Which of the following can be inferred from this information (choose all that apply)?

Select one or more:

a. The p value of the model is 1.3.

b. The model is able to identify 10% of the population that is 1.3 times as likely to graduate.

c. The overall probability of graduating is .70, therefore the model can identify a subgroup that has greater than a 90% chance of graduating.

d. 1.3 is a relatively low number, therefore the model is bad.

e. For every one unit increases in the model score, the probability of graduating goes up by 1.3%

3- A model built on a training dataset and applied to test data set shows a lot of degradation. The r square on the build is .78 while on the validation it is only .22. Which of the following does this result support? This indicates that the model has high bias error.

Select one: True False

4- Which of the following is not true of decision trees (select the single best answer):

Select one:

a. Decision trees are an example of supervised learning. b. Decision trees are non-paranetric c. Decision trees always work better than linear regression.

d. Decision trees can used in regression (or prediction) problems and classification problems.

5- A company is concerned with the size of a hole being punched into a metal sheet by an automated drill press in a production process. The company intends to use several environmental factors as predictors to determine: a) the size of the hole punched; b) whether or not the hole punched was within specification (in spec or out of spec). What kind of problem is being dealt with in each case?

Select one:

a. The size of the hole is a classification problem while the in spec/out of spec model is a regression (or prediction) problem. b. The size of the hole is a regression (or prediction) problem while the in spec/out of spec model is a classification problem. c. Both are regression (or prediction) problems. d. Both are classification problems.

6 - Members of the ACME Analytics team are in the process of building a new model to predict average sales. If they add an additional variable to the model, the r square increases, but the adjusted r square decreases and the AIC and BIC increase. What does this indicate about the model with the additional variable (select only the best answer)?

Select one:

a. Even though the r square increases, the model with the additional variable is likely overfitting the data set on which it is being built (indicated by the decreasing adjusted r square. Therefore, the team should stick to the model with fewer variables.

b. In a regression problem, the objective is always to increase the r square so the team should use the model with more variables.

c. The team cannot discern any guidance based on the information given.

7- You are trying to build a model to predict whether or not a particular visitor is likely to make a purchase on your website. You have information on over a million people who have visited the site in the past twelve months. You have over 4000 variables related to each visitor and you suspect there are complex interactions between those variables and whether someone is likely to make a purchase. Which of the following is not a viable option in performing this analysis?

Select one:

a. Linear Regression

b. Decision Tree

c. Boosted Decision Tree

d. Logistic Regression