Does anyone know such? I tried to find such procedures in Gretl, but there you can use either hsk procedure for heteroskedasticity correction or ar1 procedure for serial correlation correction. I need the GLS procedure that will deal with both. Thanks
gls() in package nlme for R can handle both situations. The serial correlation is specified via argument
correlation and heteroskedasticity is specified via the
Both can take a number of pre-specified functions also provided by the package which can estimate any parameters required. For example, the serial correlation can be specified using
corCAR1() for AR(1), ARMA and continuous time AR(1) serial correlations.
AUTOBOX is a program that I am familiar with having written both of these modules for AFS a company that I am still working with. The distinguishing features of AUTOBOX is that it can automatically determine the weights required for GLS while automatically identifying the ARIMA structure. They have a 30 day free demo at http://www.autobox.com/30day.exe . If you have any questions about the output you can deal with them directly.