Multiple Regression

Class notes

Predict Exam grade based on hours studied and GPA

First Order Model: Y= Exam grade X1= Hours studied X2= GPA E(Y)= b0+b1(Hours Studied)+b2(GPA)

E (Y ) 0 1 X 1 2 X 2

Example First order

X1= GPA X2= work experience b0 = 30,000 b1 = 5000 b2 = 500

Intercept: In the population, the average value of starting salary is 30000 when both GPA and work experience are zero. (Many times the intercept does not have a realistic meaning.)
Slope of X1: In the population, for an increase of one point in GPA, the average starting salary increases by 5000 holding work experience constant. Slope of X2: In the population, for an increase of one year in work experience, the average starting salary increases by 500 holding GPA constant.

Second Order Model: The change in the average exam grade when you study an additional hour depends on how long you have studied, controlling for the effect of GPA. For the same value of GPA change in exam grade is:

Example Second order

Assuming the first variable is squared. To create new examples click on any green shaded cell and change names and values. Y = starting salary X1= GPA X2= work experience b0 = b1 = b2 = b3 = 30,000 5000 500 100

increases by 100point increases

E(y) = 0 + 1 X1 + 2 X12 + 3 X2

Continued
Intercept: In the population, the average value of starting salary is 30000 when both GPA and work experience are zero. (Many times the intercept does not have a realistic meaning.) Effect of GPA: In the population, for an increase of one point in GPA, the average starting salary increases (or decreases) by 5000 + 500 times ( 1+ twice the current value of GPA) holding work experience constant.

Effect of work experience: In the population, for an increase of one year in work experience, the average starting salary increases by 100, holding GPA constant.

Interaction Model
The change in the average exam grade when for each additional hour studied depends on the value of the GPA.

Example: Interaction Model

X1= X2= b0 = b1 = b2 = b3 = GPA work experience 30,000 5000 500 100

E(y) = 0 + 1 X1 + 2 X2 + 3 X1X2
Intercept: In the population, the average value of starting salary is 30000 when both GPA and work experience are zero. (Many times the intercept does not have a realistic meaning.)

Effect of GPA: In the population, for an increase of one point in GPA, the average starting salary increases (or decreases) by 5000 + 100 for each year of work experience.
Effect of work experience: In the population, for an increase of one year in work experience, the average starting salary increases (or decreases) by 500 + 100 for each point of GPA.