Scribd
Upload
6 views13 pages

Heteroscedasticity

Heteroscedasticity is a violation of OLS assumptions where the variance of the error term changes with the magnitude of the explanatory variable, leading to inefficient and inconsistent estimates. It can be detected through informal methods like plotting residuals or formal tests such as the Breusch-Pagan and White General tests. Remedies include using Generalized Least Squares (GLS) or White's heteroscedasticity-consistent variances and standard errors.

Uploaded by

ndubi brian
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
6 views13 pages

Heteroscedasticity

Heteroscedasticity is a violation of OLS assumptions where the variance of the error term changes with the magnitude of the explanatory variable, leading to inefficient and inconsistent estimates. It can be detected through informal methods like plotting residuals or formal tests such as the Breusch-Pagan and White General tests. Remedies include using Generalized Least Squares (GLS) or White's heteroscedasticity-consistent variances and standard errors.

Uploaded by

ndubi brian
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 13

Heteroscedasticity

• This is violation of OLS assumption where variance of the error term


changes with magnitude explanatory variable i.e E(𝜀 𝐼 𝜀)= σ2 𝑖
Causes

• Presence of outliers
• Manipulation of data in wrong way
• Omission of important variable
• Inclusion of irrelevant variable
• Poor data collection methods
Consequences

• The estimates are unbiased


• Formula for calculating variance is wrong
• The estimates are inefficient
• Estimates are inconsistent
Detection

• There are two methods


Informal
Formal
• Informal Method
Plot 𝜀 2 against X (The explanatory suspected to cause
heteroscedasticity.
If there is a consistent pattern (e.g increasing or decreasing trend)
then heteroscedasticity is said to be present
If there is no pattern then it is absent
Formal Method

• There are many methods for testing presence of heteroscedastacity.


• We shall discuss two, namely:
Breusch-Pagan test
White General test
Breusch Pagan(BP) Test

• H0: No heteroscedasticity/homoscedasticity
• Ha: Heteroscedasticity Present
𝑌𝑖 = 𝛽0 +𝛽1 𝑋1 +𝛽2 𝑋2 +𝜀𝑖
𝜀 2 = 𝛼0 +𝛼1 𝑋1 +𝛼2 𝑋2 +𝑢𝑖
• If 𝛼1 = 𝛼2 = 0, then 𝜀 2 = 𝛼0 , a constant. This is the theory behind the
Breusch-Pagan test
Procedure
• Estimate equation and obtain errors
2 𝑢𝑖2
• Obtain σ = 𝑛
𝑢𝑖2
• Construct variables 𝑝𝑖 = σ2
• Regress on explanatory variables
• 𝑝𝑖 = 𝛼0 +𝛼1 𝑋1+𝛼2 𝑋2 +𝑒𝑖
• The test statistic is a chi-squared test
1
• Obtain Chi-square Calculated ߯ 2 = 2ESS
• Obtain ߯ 2 critical with (k-1) where k is number of parameters.
• Alternatively, we can use probability (p value). If the test statistic has a p value below an appropriate
threshold (e.g p<0.05) then the null hypothesis of homoscedasticity is rejected. The Stata command is as
follows.
• Reg yx1x2x3
• estat hettest
White General Test

• 𝑌𝑖 = 𝛽0 +𝛽1 𝑋1 +𝛽2 𝑋2 +∈𝑖 ….1


• It assumes that the relationship between variance of the error and independent variable is linear or nonlinear.
• ∈2 = 𝛼0 +𝛼1 𝑋1+𝛼2 𝑋2 +𝛼3 𝑋1 𝑋2 +𝛼4 𝑋1 2 +𝑢𝑖 ………..2
• Obtain 𝑅2 from this auxiliary eqn 2
• Obtain the Lagrange Multiplier (LM) given as
• LM= n𝑅2 where n is number of observations
• Obtain ߯ 2 critical with (k-1) where k is number of parameters in eqn 2
• Alternatively, we can use probability (p value). If the test statistic has a p value below an appropriate
threshold (e.g p<0.05) then the null hypothesis of homoscedasticity is rejected.
• The Stata command is as follows.
• Reg Y X1 X2 X3
• Stata Command : lmtest, white
Remedy for Heteroscedasticity

• If heteroscedasticity is present, then one can adopt any of the


following remedies:

• Generalized Least Squares (GLS)

• White’s Heteroscedasticity-Consistent Variances and Standard Errors


(Robust Standard errors).
GLS

• Assume we are given the following model


Y = 𝛽0 +𝛽1 𝑋1 +𝜀𝑖
• Where the Var(𝜀) = 𝛔2 𝑋12
• This implies that the variance is not constant but varies with 𝑋1 , a
problem of heteroscedasticity
• GLS is used in such case. It transforms the model in such a way that
the Var(𝜀) = 𝛔2 , a constant. That is homoscedastic error term.
1 2 2
Var(𝜀) = 2 𝛔 𝑋1 = 𝛔2 (a constant)
𝑋1
GLS Cont

• Step 1: Transform the model by dividing throughout by 𝑋1 as follows.


Y = 𝛽0 +𝛽1 𝑋1 +𝜀
Y 𝛽0 𝛽1 𝑋1 𝜀
= + +
𝑋1 𝑋1 𝑋1 𝑋1
1
𝑌∗ = 𝛽1 +𝛽0 + 𝑉
𝑋1
𝑌 ∗ = 𝛽1 +𝛽0 𝑋 ∗ + 𝑉 ……..This is the transformed model
∗ Y ∗ 1 𝜀
• Where 𝑌 = , 𝑋 = , V =
𝑋1 𝑋1 𝑋1
GLS Cont

• Comparing OLS and GLS


OLS : 𝛽𝑂𝐿𝑆 = 𝑋 𝐼 𝑋 −1 𝑋 𝐼 Y
GLS : 𝛽𝐺𝐿𝑆 = 𝑋 𝐼 𝛀𝑋 −1 𝑋 𝐼 𝛀Y
1
• Where Omega(𝛀) =
𝑋1 −1
𝑁 𝑋1∗ 𝑌∗
𝛽1𝐺𝐿𝑆
=
𝛽0𝐺𝐿𝑆 ∗2
𝑋1∗ 𝑋1 𝑌∗𝑋∗
White’s Heteroscedasticity-Consistent Variances and
Standard Errors.
• Several computer packages present White’s heteroscedasticity-
corrected variances and standard errors along with the usual OLS
variances and standard errors.
• Incidentally, White’s heteroscedasticity-corrected standard errors are
also known as robust standard errors.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy