Table 1 Variance inflation factor (VIF) tests to identify and remove variables with multicollinearity

From: Multilayer networks characterize human-mobility patterns by industry sector for the 2021 Texas winter storm

Category

Predictor

Test 1

Test 2

Test 3

Test 4

Demographic

Population

5.318

4.785

3.531

3.375

 

Population Density

3.308

3.292

2.687

2.672

 

65 years and over (%)

8.458

5.516

5.444

–

 

Under 18 (%)

84.755

–

–

–

Socioeconomic

Income

14.488

6.156

5.177

2.675

 

School Enrollment (%)

63.847

12.666

–

–

 

Unemployment (%)

4.884

4.814

4.535

4.255

 

Poverty Rate (%)

8.111

7.557

4.021

3.933

 

Renter Occupied (%)

18.919

8.367

–

–

 

Owner Occupied (%)

28.744

–

–

–

Race/Ethnicity

Non-White (%)

6.201

5.609

5.038

4.849

 

Non-Hispanic and Non-Black (%)

10.992

–

–

–

  1. For Test 1, we used 12 predictors from three categories, resulting in most VIF values being greater than 5 (which is a standard threshold). Then we conducted successive tests to remove the highest VIF score from each category (while leaving at least one per category). For the final test (Test 4), each predictor in Test 3 was removed one at a time to identify a set of factors such that no VIF values were greater than 5.
Back to article page