Hi everyone,
I would like to seek help me with some maths:
BACKGROUND
The individual deviance residuals d_i is defined as:
d_i = sign(y_i - mu_i)*sqrt{2*[S - C])
where
S = log likelihood of saturated model, the data y_i.
C = log likelihood of current model mu_i.
and DEVIANCE = sum over all i of (d_i)^2
I am using Stata 6.0 software fitting a negative binomial GEE on repeated
measures cell counts.
It reports the DEVIANCE, but I cannot get the individual d_i for model
diagnostics.
QUESTION
Can anyone provide me the formula for calculating d_i for negative binomial
glm.
(Note: I have been able to do it for Binomial and Poisson glm, but not for
negtive binomial.)
Thank you.
Tri
________________
Tri Tat
Medical Statistician
Tel: 020 7351 8697
Fax: 020 7351 8578
Research Services
Statistical Support
Royal Brompton &
Harefield NHS Trust
Chelsea Wing Level 2
Privileged/Confidential Information and/or Copyright Material may be
contained in this e-mail.
The information and material is intended for the use of the intended
addressee, or the person responsible for delivering it to the intended
addressee, you may not copy or deliver it to anyone else or use it in any
unauthorised manner.
To do so is prohibited and may be unlawful.
If you receive this email by mistake, please advise the sender immediately
by using the reply facility in your e-mail software, or notify the Royal
Brompton & Harefield NHS Trust Helpdesk on +44 (0)20 7351 8696.
Communication is not sent through a secure server; Royal Brompton &
Harefield NHS Trust cannot accept responsibility for the accuracy of
outgoing electronic mail.
Any views or opinions presented are solely those of the author and do not
represent the views of the Royal Brompton & Harefield NHS Trust unless
specifically stated.
|