Dear All,
I need a small help on dispersion parameter
interpretation in glm; well glm fixes the dispersion
parameter to be 1.
For example I am fitting log-linear model with
family=poisson. To check the overdispersion parameter
can I fit the model with family=quasi(link=log,
variance=mu);
Q1. Does family=quasi(link=log, variance=mu) will be
equivalent to family=poisson?
Q2. If they what does dispersion parameter indicate
in the quasi for quasi(link=log, variance=mu)?
Q3. if the dispersion parameter is wide away from 1
in quasi fit, then what we conclude about the fit.
can you refer me to any specific chapter of any book.
Many thanks,
Jag
__________________________________
Do you Yahoo!?
Get better spam protection with Yahoo! Mail.
http://antispam.yahoo.com/tools
|