Multiple Regression

Question 1

Logic:

Answer 1

Is there reason to expect a causal relationship between the predictors and the response variable.
• Like Correlation, Regression can’t prove causation – that assumption is on us as the data collectors/owners.

Question 2

Fit:

Answer 2

Does the overall regression show a significant relationship between the predictors and the response variable?

Question 3

Parsimony:

Answer 3

Does each predictor contribute significantly to the explanation? Are some predictors not worth the trouble?

Question 4

Stability:

Answer 4

Are the predictors related to one another so strongly that regression estimates become erratic?

Question 5

Occam’s Razor

Answer 5

When two explanations are otherwise equivalent, we prefer the simpler one

Question 6

R2 Adjusted

Answer 6

• The closer R2 adjusted is to R2 the more each predictor is contributing
• The further apart the weaker the contributors
• R2 Adjusted will always be less than R2