On 14 May 2002 11:15:14 -0700, [EMAIL PROTECTED] (Brian Leung)
wrote:

> I am fitting a binary response variable to the Weibull function
> 
> y=1-exp(-x/b)^c
> 
> I am using maximum likelihood techniques to determine b and c. Given
> my estimate of c, I want to determine what the probability is that the
> true value of c<=1.It appears that the likelihood ratio may allow me
> to find this probability. However, for binary response variables, the
> likelihood ratio is apparently biased. Does anyone have a suggestion?

What is biased?  the LR solution for c?
 - toward what, away from what, and when?
 - is the bias known?  Is it large?

 - a suggestion for ....   Does it look like you can't trust
the p-level about your answer for c?

Do you have a reference that suggests your problem
is serious, or just how serious the problem might be?

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to