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

From |
"Marc F. Bellemare" <mfb24@cornell.edu> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: RE: Log-Likelihood of Each Observation |

Date |
Sat, 09 Oct 2004 17:19:23 -0400 |

Dear colleagues:

Stas was pretty close. I tried his suggestion, and my ml program was not able to find feasible initial values anymore! (This strikes me as rather odd, by the way...) So what I did was to generate (prior to defining my ml program) a new variable called logL, and to add at the end of my ml program which is just "replace logL = `lnf'. It now works just fine.

To answer Stas' question, I am running a Vuong test to discriminate between two non-nested models, one of which is a bivariate probit, in order to see which is best supported by my data.

Thanks very much, Stas' and Maarten!

Marc

At 11:04 PM 10/9/2004 +0200, you wrote:

Dear mark:

You would not generate a variable (althought you could if you really wanted to). After any estimation command a number of statistics are temporarily stored. You can see which by looking in the manual or type -ereturn list-. In the example below you can see that the log likelihood is stored in e(ll). You can use that in your calculations. See -help return- and the manual entry for return for more information.

Example:

. sysuse auto

(1978 Automobile Data)

. quietly logit foreign weight mpg

. ereturn list

scalars:

e(N) = 74

e(ll_0) = -45.0332095569914

e(ll) = -27.17515619914122

e(df_m) = 2

e(chi2) = 35.71610671570034

e(r2_p) = .3965529779806182

macros:

e(depvar) : "foreign"

e(cmd) : "logit"

e(crittype) : "log likelihood"

e(predict) : "logit_p"

e(chi2type) : "LR"

matrices:

e(b) : 1 x 3

e(V) : 3 x 3

functions:

e(sample)

. di e(ll)

-27.175156

Maarten

-----Original Message-----

From: Marc F. Bellemare [mailto:mfb24@cornell.edu]

Sent: Sat 10/9/2004 10:44 PM

To: statalist@hsphsun2.harvard.edu

Cc:

Subject: st: Log-Likelihood of Each Observation

Dear colleagues:

I was wondering if anyone knew how to generate a variable equal to the

log-likelihood of a model just evaluated (either when using the ml subset

of command or not). I would need such a variable to compute the test

statistic for a Vuong (1989) test.

Cheers,

Marc

__________

Marc F. Bellemare

Ph.D. Candidate

Department of Applied Economics and Management

Cornell University

Ithaca, NY 14853-7801

http://www.people.cornell.edu/pages/mfb24

*

* 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/

__________ Marc F. Bellemare Ph.D. Candidate Department of Applied Economics and Management Cornell University Ithaca, NY 14853-7801 http://www.people.cornell.edu/pages/mfb24 * * 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/

**References**:**st: RE: Log-Likelihood of Each Observation***From:*"Maarten Buis" <M.Buis@fsw.vu.nl>

- Prev by Date:
**st: RE: Log-Likelihood of Each Observation** - Next by Date:
**Re: st: Problem of convergence** - Previous by thread:
**st: RE: Log-Likelihood of Each Observation** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |