Detecting and dealing with multicollinearity

Question 1

What is the first way of detecting multicollinearity?

Answer 1

Look at R Squared and t ratios 
High R(sq) but few significant t ratios = multicollinearity because it affects the size of the t values by making them generally smaller

Question 2

What is the second way of detecting multicollinearity?

Answer 2

calculate the correlation between independent variables.

High pair-wise correlations between independent variables at (70+) are signs of multicollinearity

Question 3

What is the third way of detecting multicollinearity?

Answer 3

To detect linear relasionships between 2+ indepdnant variables run a t test regressions.

I.E regression analysis performed between each of the indepdnant variables and the remiander

examine R(sq)

Question 4

What is the first way of dealing with multicollinearity?

Answer 4

delte highly correlated variables

if variables are highly correlated, keep 1 in the model

Question 5

What is the second way of dealing with multicollinearity?

Answer 5

increase the sample size

more obsevations increase ∑X2j in (3) and (4)

therefore increasing the variability in the regressors we can reduce the collinearity between them