ECON61001 Econometric Methods Ekaterina Kazak Computer Tutorial 6, page 1 of 1 University of Manchester
The purpose of this session is to introduce you to the heteroscedastic error term and different tests for heteroscedasticity.
Heteroscedasticity
Use the Wooldridge dataset hprice1 to obtain the usual OLS and heteroskedastic-robust standard errors for the following model of house prices,
price = ¦Â0 + ¦Â1 lotsize + ¦Â2 sqtft + ¦Â3 bdrms + u (1)
a) Discuss any important differences between the usual OLS and heteroskedastic-robust standard
errors.
b) Repeat part (a) for the following model,
log(price) = ¦Â0 + ¦Â1 log(lotsize) + ¦Â2 log(sqtft) + ¦Â3 bdrms + u (2)
c) What does this example suggest about the heteroskedasticity present in model (1) and the
transformation used for the dependent variable.
d) Consider model (1). Conduct a Breusch-Pagan test, a White test, and the special case of the White test using both the LM and F forms of the appropriate test statistics. Complete the following table,
Test
Observed test statistic
Critical value for 5% test
p-value
Reject(Y/N)
F
LM
F
LM
F
LM
F
LM
BP
White
White (Special)
Hint: Generate and save the residuals using the resid commands. Also generate new variables for the squared values and interaction terms, then run the required regressions to obtain the necessary test statistics. The following command, qchisq(x,df) will provide the p-value for the ¦Ö2 distribution with df degrees of freedom and an observed test statistic of x.