[Corpora-List] variant log likelihood calculations

Rayson, Paul rayson at exchange.lancs.ac.uk
Wed Dec 15 11:36:01 CET 2004


Dear Don,

Glad you figured out the problem. Had me worried there for a moment!

The version of the formula I use comes from the Cressie and Read paper(s) that we reference in the publications you listed. For more details, an on-line LL calculator, and the papers, see

http://ucrel.lancs.ac.uk/llwizard.html

Note that ln(0) is undefined, so I pre-define it to be zero. Another approach might be to use a very small value estimate for words with zero frequency.

Regards,
Paul.

Dr. Paul Rayson
Director of UCREL (University Centre for Computer Corpus Research on Language)
Computing Department, Infolab21, South Drive, Lancaster University, Lancaster, LA1 4WA, UK.
Web: http://www.comp.lancs.ac.uk/computing/users/paul/
New telephone number: +44 1524 510357 Fax: +44 1524 510492


-----Original Message-----
From: owner-corpora at lists.uib.no [mailto:owner-corpora at lists.uib.no]On
Behalf Of Don Hardy
Sent: 15 December 2004 05:41
To: Don Hardy
Cc: CORPORA
Subject: Re: [Corpora-List] variant log likelihood calculations



I just figured out what I was doing wrong. I wasn't carrying the Rayson
and Garside calculation through for all cells of the contingency table.

Thanks for the help.

Don







More information about the Corpora-archive mailing list