Path: news.mathworks.com!not-for-mail
From: Ting <ting_noreply@mathworks.com>
Newsgroups: comp.soft-sys.matlab
Subject: Re: Log-likelihood
Date: Thu, 16 Oct 2008 10:59:37 -0400
Organization: The MathWorks, Inc.
Lines: 12
Message-ID: <gd7ksp$30o$1@fred.mathworks.com>
References: <gd7j91$cv3$1@fred.mathworks.com>
NNTP-Posting-Host: sut.dhcp.mathworks.com
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
X-Trace: fred.mathworks.com 1224169177 3096 172.31.57.42 (16 Oct 2008 14:59:37 GMT)
X-Complaints-To: news@mathworks.com
NNTP-Posting-Date: Thu, 16 Oct 2008 14:59:37 +0000 (UTC)
User-Agent: Thunderbird 2.0.0.17 (Windows/20080914)
In-Reply-To: <gd7j91$cv3$1@fred.mathworks.com>
Xref: news.mathworks.com comp.soft-sys.matlab:495630


Thiago,
  People usually use maximum likelihood method to estimate parameters 
for   a chosen model, but You should not use log-likelihood to compare 
different models. Because, generally a model with more parameters tends 
to have larger likelihood.

There is no simple answer for how to decide the goodness of fit for 
distribution fit. For details, please see my reply for your question on 
"Doing Goodness of Fit test in dfittool'.

-Ting Su
Mathworks.