Scribd
Upload
62 views

Homoscedasticity, Heteroscedasticity and Multicollinearity

This document discusses tests for homoscedasticity and heteroscedasticity. It describes the Park test, Glejser test, and White test to detect heteroscedasticity. It also discusses detecting multicollinearity using correlation, coefficient variance decomposition, condition numbers, and variance inflation factors. Tests for heteroscedasticity include regressing the residuals on the independent variables or their absolute values and seeing if the coefficients are statistically significant. High correlation between variables, condition numbers over 30, and VIF scores above 10 indicate multicollinearity.

Uploaded by

Nurul Ariffah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
62 views

Homoscedasticity, Heteroscedasticity and Multicollinearity

This document discusses tests for homoscedasticity and heteroscedasticity. It describes the Park test, Glejser test, and White test to detect heteroscedasticity. It also discusses detecting multicollinearity using correlation, coefficient variance decomposition, condition numbers, and variance inflation factors. Tests for heteroscedasticity include regressing the residuals on the independent variables or their absolute values and seeing if the coefficients are statistically significant. High correlation between variables, condition numbers over 30, and VIF scores above 10 indicate multicollinearity.

Uploaded by

Nurul Ariffah
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

Homoscedasticity and Heteroscedasticity.

1. Park test: (cannot reject H0 = Homoscedasticity)


 Create OLS: Quick > estimate equation Ho= Homoscedasticity
H1= Heteroscedasticity

Prob < 0.01, 0.05, 0.10


Significant = reject H0 but accept H1

Prob > 0.01, 0.05, 0.10


Insignificant = accept/cannot reject H0 but
reject H1
 Make residual: View > Actual, Fitted, Residual > Actual, Fitted, Residual Table.

Copy all the residual value.

 Generate series for log/ln variables: Quick > Generate series.

Generate series:
Lncpi=log(cpi)
Lnipi=log(ipi)
Lngold=log(gold)
 Estimate equation: Quick > estimate equation.

Insignificant.
Cannot reject H0
2. Glejser test :
 Convert to absolute value.
 To detect Heteroscedasticity.
 Generate series for absolute value: Quick > Generate series.

Negative value to positive value.


 Estimate equation for Inverse, Squared and Actual.

Significant. Insignificant.

Reject H0 Cannot reject H0

Significant.

Reject H0
 Using Heteroscedasticity test.

Significant or not?
3. White test.

Plot scatter graph.

Generate series at the command for ipi u and cpi u.

 Using Heteroscedasticity test > white.


Multicollinearity.

 Correlation.

Estimate eq:
gold c cpi ipi

The highest value have correlation.


 Coefficient Variance Decomposition.

CDV: k between 100 to 1000 =


Multicollinearity.
 Condition scalar:

Multicollinearity.

 Variance Inflation Factors.

VIF > 10

Multicollinearity.

You might also like