Econometrics Lab Notes
Uploaded by
i234519Econometrics Lab Notes
Uploaded by
i234519ECONOMETRICS LAB NOTES
Lab task 5: VIF
🧮 In Excel
Steps
1. Choose variables:
o Suppose you have three independent variables: X1, X2, and
X3.
o Treat one variable as dependent (D.V) and the remaining
two as independents ([Link]).
o Repeat this for all variables:
Regression 1: X1 = f(X2, X3)
Regression 2: X2 = f(X1, X3)
Regression 3: X3 = f(X1, X2)
2. Run Regression:
o Go to Data → Data Analysis → Regression.
o Select your D.V and [Link] accordingly.
o Record the R² value from each regression output.
3. Calculate VIF:
o Use the formula in Excel:
o =1/(1-R^2)
o Compute for each variable.
4. Interpretation:
o VIF < 5 → Acceptable
o VIF > 10 → High multicollinearity (problematic)
📈 In EViews
Steps
1. Select Variables:
o Highlight the Dependent Variable (D.V) and Independent
Variables ([Link]).
2. Run Regression:
o Right-click → Open as Equation.
o Click OK to run the regression.
3. Run VIF Test:
o In the Equation window → go to View → Coefficient
Diagnostics → Variance Inflation Factors (VIF).
4. Interpret Results:
o VIF values will appear for each independent variable.
o Interpret as before (VIF < 5 = good, >10 = problematic).
📘 In SPSS
Steps
1. Import Data:
o Open SPSS → File → Import Data → Excel/CSV.
2. Run Regression:
o Go to Analyze → Regression → Linear.
3. Select Variables:
o Move your Dependent Variable to the Dependent box.
o Move your Independent Variables to the Independent(s)
box.
4. Enable Collinearity Diagnostics:
o Click Statistics → check “Collinearity diagnostics” →
Continue → OK.
5. View Output:
o In the Coefficients table, see VIF under the “Collinearity
Statistics” column.
Lab Task 6
Step 1 — Run the main regression
Using Data Analysis → Regression, run your model and store the
residuals.
Step 2 — Create residual squared
In a new column:
e2 = Residual^2
This is required for the Breusch–Pagan test.
Step 3 — Run auxiliary regression
Run another regression:
Dependent Variable: e²
Independent Variables: all original X variables
From this regression, note:
R² (auxiliary)
Step 4 — Compute the LM statistic
Formula:
2
LM =n × R aux
Where n = sample size.
Step 5 — Compare LM with Chi-Square
Degrees of freedom = number of regressors tested (k)
If LM > χ 2critical → heteroskedasticity present
Or check p-value if you computed it:
2
p=1− χ CDF( LM , k )
If p < 0.05, heteroskedasticity exists.
✔ Your method is correct
You took residuals → squared → auxiliary regression → compared LM with
chi-square → that is exactly the Breusch–Pagan test.
✅ 2. How to Test for Heteroskedasticity in EViews
EViews has built-in heteroskedasticity tests, so the process is much
simpler.
Method A — Breusch–Pagan Test
After running your regression:
1. Open the Equation window
2. Go to View → Residual Diagnostics → Heteroskedasticity Tests
3. Select:
o Breusch–Pagan–Godfrey
o OR White test
EViews will automatically show:
LM statistic
F statistic
p-values
Decision (significance)
Interpreting results:
p < 0.05 → heteroskedasticity present
p > 0.05 → no heteroskedasticity
✅ 3. How to Test for Heteroskedasticity in SPSS
SPSS does not have built-in heteroskedasticity tests like BP or White, but
you can do the manual BP test, similar to Excel.
Method A — Glejser Test (Standard for SPSS)
SPSS users commonly use the Glejser test, which is equivalent in
purpose.
Steps:
1. Run the main regression (Analyze → Regression → Linear)
2. Save unstandardized residuals
o Click Save
o Tick Residuals → Unstandardized
3. Compute new variables:
o ABSRES = ABS(residuals)
4. Run regression:
o Dependent variable: ABSRES
o Independent variables: all original X variables
✔ Interpretation:
If any X has significant coefficients (p < 0.05) → heteroskedasticity
exists.
Method B — Manual Breusch–Pagan Test in SPSS
If you want the exact BP test:
Steps:
1. Save Unstandardized Residuals
2. Compute:
3. e2 = residual^2
4. Run new regression:
o DV: e2
o IVs: original Xs
5. Note R²
6. Compute:
2
LM =n × R
7. Compare LM with Chi-Square critical (df = k)