Are the values of these points input correctly
Remove one of these predictors whose VIF are greater than 5 and make a new regression model
Are there any discernible patterns in the plots
Whether at least one predictor in this subset is linearly related to the response
Whether condition 1 and 2 are both met
Yes
Delete or correct this point
No
Acess assumptions with residual plots and QQ-plot
Remove this subset of predictors and do a linear regression again.
Srart with the biggest model I can bulid with the predictors
Is there any predictor whose corresponding VIF value greater than 5
Use Box-Cox transformations on the model
B
Use VIF method to test whether multicollinearity exists in each remaining predictor
Record these points and get the final model
Draw scatterplots and the plot of the response against the fitted values to check condition 1 and 2
Remove a predictor from this subset
