LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (March 2006, week 1)Back to main SAS-L pageJoin or leave SAS-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:   Thu, 2 Mar 2006 22:14:07 -0500
Reply-To:   Tony Yang <tonyyangsxz@GMAIL.COM>
Sender:   "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:   Tony Yang <tonyyangsxz@GMAIL.COM>
Subject:   Re: proc genmod-criteria for overdispersion
Comments:   To: nevin.krishna@gmail.com
In-Reply-To:   <f00d287f0603021912v38903299ic955be2ba8203e6d@mail.gmail.com>
Content-Type:   text/plain; charset=ISO-8859-1

> Hi, Nevin; > > The value 1.3935 is hard to say if the Poisson model has overdispersion, > while you can try an alternative model such as negative binomial or > generalized Poisson model to fit the data, and you can try to use the score > test if the NB or GP model is better than the Poisson model. > > Meanwhile, there is a dispersion index, defined as d=variance/mean, > if d>1, we can consider there is overdispersion in your data, this time you > can try NB or GP model, since 1.3935 is not big enough to consider other > models. > > Also, I suggest you make a plot for your response count variable, the > x-axis is the count, and the y-axis is the observed percent, to see if there > is some characteristic. Generally speaking, Poisson model, NB and GP model > will fit a unimodal data, if your data do not have this characteristic, you > may consider other alternatives. > > I am attaching my submitted paper for your reference. > > > On 3/2/06, nevin.krishna@gmail.com <nevin.krishna@gmail.com> wrote: > > > > Hello all, > > > > i am trying to do a poisson regression with an outcome being rate of > > disease, and independent variables being agecode (categories of age), > > region, and sex. > > I am trying to figure out how to interpret whether there is evidence of > > overdispersion: i was taught that this can be accomplished by > > looking at the value/df from the criteria for assessing goodness of fit > > output...the closer the value is to 1 the less the influence of > > overdispersion. > > > > when i run the following code, i get the following output.. > > > > > > proc genmod data=mening_poisson; > > class agecode sex region; > > model count=agecode sex region / offset=l_pop dist=poi link=log > > type3; > > run; > > > > Criteria For Assessing Goodness Of Fit > > > > Criterion DF Value > > Value/DF > > > > Deviance 161 224.3462 > > 1.3935 > > Scaled Deviance 161 224.3462 > > 1.3935 > > Pearson Chi-Square 161 204.5569 > > 1.2705 > > Scaled Pearson X2 161 204.5569 > > 1.2705 > > Log Likelihood 843.6215 > > > > > > Is 1.39 considered close to 1 ? what is the cutpoint for what > > determines overdispersion based on the value/df stat? > > > > Thanks, nevin > > > > > > -- > Best regards, > > Tony > >

-- Best regards, Tony


Back to: Top of message | Previous page | Main SAS-L page