[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
"Mindaugas Baltutis" <[email protected]> |

To |
<[email protected]> |

Subject |
st: -ml- estimators for time series (keys: ts, ml, arch) |

Date |
Tue, 13 Dec 2005 15:22:47 +0100 |

Dear Stata Users, I expect to write various Multivariate GARCH estimators on Stata, but at the moment I am even stuck at writing my own estimator for simple GARCH(1,1) model using -ml-. I am a newbie to Stata, so I apologize if my questions are naï¿½ve. 1. The info about -ml- which I found in "[R] ml" was not enough for me, because no examples were given how to specify time series estimators. In particular: 2. Which -ml- method {lf,d0,d1,..} is appropriate for ts likelihood like in garch. In this case it is a conditional likelihood, and that is a problem: I don't find the way to fit in the framework of -ml-, i.e. how to specify the `theta' when it depends on its own lags. I do it straightforward: ........ program myarch_dr version 8 args lnf sigma2 qui replace `lnf' = -0.5 * (ln(2*_pi) + ln(`sigma2') + $ML_y1 /`sigma2') end cmd: ml model lf myarch_dr ( e2_t = L.e2_t L.sigma2) in 2/l, collinear ........ If all `sigma2' values are initialized then the model is estimated, but with these static values. And if I supply only the initial value for `sigma2', with rest missing, the output is "insufficient observation". 3. I did examine the own Stata's arch.ado file. And there is a different approach and commands, for which I didn't find help. In particular: The call for -ml- method is "rdu0". What is that? ......... capture noi ml model rdu0 `ml_prog' `xbeqn' .... etc.. ......... And the log-likelihood is calculated in observation loop for which I also didn't find the help: ........ /* Observartion loop */ _byobs { ...... /* ARCH components */ $Tdoarch score $Tsigma2 = `b', eq("ARCH") missval(`sig2_0') `dohet' update $Tsigma2 = $Tsigma2 `hetterm' ...... /* Likelihood */ update `llvar' = -0.5 * (ln(2*_pi) + /* */ ln($Tsigma2) + $Te2 / $Tsigma2) } ........ How does this loop work? I didn't find info about the "update" and "_byobs" commands? 4. Does somebody know if the book Maximum Likelihood Estimation with Stata (Gould and Sribney 1999) Is describing these features about -ml- programming, does it have examples with time series. Thank you in advance for comments and help. Best regards, Mindaugas --- Mindaugas Baltutis Department of Finance University of Vienna Bruenner Strasse 72 A-1210 Vienna Austria Tel: +43 (0)1 4277-38076 Fax: +43 (0)1 4277-38054 * * For searches and help try: * http://www.stata.com/support/faqs/res/findit.html * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

- Prev by Date:
**st: show distribution @ top of graph** - Next by Date:
**st: RE: RE: Generating unique admission id** - Previous by thread:
**st: show distribution @ top of graph** - Next by thread:
**st: gr 2way scatter questions on how to control for specific graph appearances** - Index(es):

© Copyright 1996–2024 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |