QUESTION 1 30 MARKS

You have been brought in as a Data Science consultant on a court case. A chemical company

has been found negligent after a chemical spill at one of their plants. All that remains in the

court case is to decide on the extent of the damages for which the company is liable. One way the court has been deciding on this amount is to look at the impact the spill has had on the

value of houses located near to the chemical plant where the spill occurred.

As the expert witness, you have been asked to evaluate some alternative strategies to

estimate the impact on housing prices (price). Strategy A involves taking a sample of sales   that occurred after the spill where the houses are classified as either being close to the plant or not. This feature was designated by a variable near that was equal to 1 if the house was   deemed to be close to the chemical plant and zero otherwise. Then a regression analysis is  performed using the following model (MA):

MA: PTicei  =  β0  +  β1neaTi  +  ui.

Strategy B involves taking a sample of sales for houses near to the plant but where some   sales occurred before the spill and some after.  The variable after is equal to 1 if the house was sold after the spill and zero if the sale was before. Then a regression analysis is performed using the following model (MB):

MB: PTicei  =  β0  +  β1 afteTi  +  ui.

Part A. (5 Marks)

Explain A and B as strategies to estimate the impact of the chemical spill and critically evaluate each of them. Is either preferable to the other?

Part B. (5 Marks)

Suggest an alternative regression model that is preferable to MA given that you only have data from after the spill. Does this address all your criticisms of Strategy A that you outlined in part   (a)?

Part C. (10 Marks)

Using housing data models MA and MB are estimated, and the results given below. How do you interpret these results? (Note that PTice is expressed in $1000)

=  131.9 − 40.0neaT

(4.0)       (7.6)

n = 142, R2  = .165, StandaTdeTTOTS in ( . )

= 63.7 +  28.3afteT

(5.9)        (9. 1)

n = 96, R2  = .094, StandaTdeTTOTS in ( . )

Part D. (10 Marks)

Suppose you have sales both near and not near to the plant as well as sales before and after  the spill. Suggest an alternative strategy to estimate the effect of the oil spill on housing prices that is preferable to both MA and MB?

QUESTION 2 40 MARKS

QUESTION 3 30 MARKS

University students in a large class have access to all subject material online. To evaluate

whether students perform better in an end-of-term assignment if they review these materials prior to the due date of the assignment, consider the following model relating assignmaTK  = the assignment mark (out of 20) to a binary indicator  view  =  1 if the student viewed the material online at least once:

assignmaTK = β0  + β1 view + u.