Date:         Sat, 4 Jul 2009 13:19:59 -0400
Reply-To:     Wensui Liu <liuwensui@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Wensui Liu <liuwensui@GMAIL.COM>
Subject:      Re: Zero-inflated Poisson Regression
Comments: To: Newstat <chunyuanfei@gmail.com>
In-Reply-To:  <94c8fbb0-e46a-41b3-9e9c-eb8bd60fb43f@x6g2000prc.googlegroups.com>
Content-Type: text/plain; charset=ISO-8859-1

take a look at my sugi paper in 2008 "count data models in sas" and
you should be able to find the answer.

On Sat, Jul 4, 2009 at 3:01 AM, Newstat<chunyuanfei@gmail.com> wrote:
> I am using zero-inflated poisson regression to do data analysis. The
> outcome is number of hospitalization. 70% have never been
> hospitalized, 18% were hospitalized once, and 5% twice. The follow-up
> period (age) varied among the subjects. I have several questions.
>
> 1. the estimates in the Poisson part were not robust even when I
> slightly changed the categories of some covariants, for example, two
> age groups (<25 and 25-29 years old) were combined.
>
> 2. When I compared with the point estimates from overdispersed, they
> were even in the opposite direction.
>
> I am quite confused about which model I shall choose. Could you please
> help me figure it out? Your response is highly appreciated.
>
> Vivian
>



--
==============================
WenSui Liu
Blog   : statcompute.spaces.live.com
Tough Times Never Last. But Tough People Do.  - Robert Schuller
==============================