<html><head><style type="text/css"><!-- DIV {margin:0px;} --></style></head><body><div style="font-family:arial,helvetica,sans-serif;font-size:10pt"><div>Hi,<br><br>I added the option to estimate a garch model by the method of Quasi-Maximum Likelihood (QML) to the fGarch package. The QML assumes normal distribution and uses robust standard errors for inference. Hence, my contribution boils down to the computation of robust standard errors.<br>I made a couple of small changes to the files 'garchFit.R' and 'methods-summary.R'. I am sending you the files in the attachment because I think it is more convenient than listing the changes here in the email. I am also sending you the file 'garchGradient.R', which is used in the computation of robust standard errors. I am enclosing the corresponding manual files as well. All changes were made to the latest revision of fGarch, i.e. 3486.<br>I checked accuracy of the robust standard errors computation with the help
of the econometric software TSP. First, I simulated two garch processes and estimated them in R, see the files 'garch QMLE test.R' and 'garch QMLE test output.txt' in the attachment. Then I estimated the same models using the same data in TSP, see the file 'QMLE benchmark.out'. As can be seen in the files, the results are the same. Therefore, I believe the robust standard errors computation is accurate.<br>If you have any questions regarding the QML method, the enclosed files or the changes I made, please do not hesitate to contact me.<br><br>Best regards,<br><br>Michal<br></div></div><br>
</body></html>