MGSC 372 Lecture Notes - Lecture 5: Linear Regression, Multicollinearity, Correlation And Dependence

• School: McGill University
• Department: Management Science
• Course: MGSC 372
• Professor: Brian Smith

Document Summary

Model 4: profit on advert, special, size, place. In model 4, none of the four variables has a significant t-value. Also the increase in (cid:2870) over model 2 or multicollinearity and has an (cid:2870) value greater than 0. 90. Although the f-test is significant, this model has little to commend it over model 2 which is free of. Using the principle of parsimony (keep it simple), model 2 is preferable to model 4. Looking back at the simple correlation matrix we see that the independent variable that has the strongest linear relationship with is variable (cid:2871) (size). As we see from the next slide, the estimated regression equation is: We observe that this model has an (cid:2870) value of 0. 8363, so that variable (cid:2871) alone accounts for approximately 84% of variation in . This model has (cid:2870)=(cid:882). (cid:891)(cid:883)(cid:887)(cid:882) so that (cid:2871) and (cid:2872) together explain 91. 5% of variation in .