How to obtain Adjusted R-squared value for multiplicative heteroskedastic regression model?

[Laiy Kho]

Hello,

I employed a multiplicative heteroscedasticity model proposed by Harvey (1976) using hetregress command (two step GLS estimation). However, I am unable to obtain the adjusted r squared value for the model. It only provides these stats as attached below.

Heteroskedastic linear regression	Number of obs	=	950,000
Two-step GLS estimation
Wald chi2(23)	=	6.29e+06
Prob > chi2	=	0.0000

A research study estimated this model and has reported different adjusted r squared values for mean and variance equation (attached below). Could anyone kindly advise how I can obtain the same?

[Andrew Musau]

I would say, don't bother... not everything you see people doing makes sense. Stata is correct not to report an \(R^2\) statistic as estimation is by maximum likelihood or two-step GLS and not OLS where the statistic is most appropriate. In cases where \(R^2\) is not reported, you may generate the predicted outcome and square the correlation between the outcome and the predicted outcome, as shown in https://www.stata.com/support/faqs/s...ics/r-squared/, then adjust this to get the Adjusted \(R^2\). But as I said, I do not advise that.

[Laiy Kho]

Which statistic would be appropriate to report as an alternative for R squared value in this model?

[Carlo Lazzaro]

Laiy:
see

e(chi2) chi-squared for mean model test
e(chi2_c) chi-squared for heteroskedasticity test

from -hetregress- helpfile.