Multiple Linear Regression

Section 2: The multiple linear regression model

• Fit a preliminary model. At this point, do not discuss the regression output.

• Check model assumptions (e.g., constant variance, normality, and uncorrelated errors if

relevant) and diagnostics (outliers, leverage, influence, variance inflation); conduct

Modified-Levene test, test for normality, and Bonferroni outlier test.

IF you have adequate reason, you may remove outliers and re-do the preliminary

analysis.

• Perform necessary transformations if necessary (check online) and present the

transformed model (remember to re-check model assumptions).

• Clearly present your preliminary model that satisfies the model assumptions.

Section 3: Explore the interaction terms

• Explore interactions using partial regression plots.

• Discuss the addition of possibly useful interaction terms.

• Check correlations involving the added interaction terms before and after standardization.

(Optional)

Section 4: Model search

• Obtain a set of two potentially good models (backwards deletion & stepwise regression):

• Make sure all predictors are significant at the ???? = 0.10 level, and multicollinearity is not a

serious problem.

• Clearly present your potentially good models.