Date:         Wed, 10 May 2006 11:57:12 -0500
Reply-To:     baogong jiang <bgjiang@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         baogong jiang <bgjiang@GMAIL.COM>
Subject:      Re: fit a ZINB model using PROC NLMIXED
Comments: To: Dale McLerran <stringplayer_2@yahoo.com>
In-Reply-To:  <20060510164443.20913.qmail@web32208.mail.mud.yahoo.com>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Dale:

Thank you a lot. yes, it did work after I initializing the parameters
to zero


baogong


On 5/10/06, Dale McLerran <stringplayer_2@yahoo.com> wrote:
>
> --- baogong jiang <bgjiang@GMAIL.COM> wrote:
>
> > Dear SAS-Lers:
> >
> > I am trying to fit a ZINB model in proc nlmixed:
> >
> > I submit the following code:
> >
> > *
> >
> > proc* *nlmixed* data=srcdata.bj_04_compliance_final_5;
> >
> > eta_zip = b0_zip + b1_zip * charlson + b2_zip * no_of_diag +
> >
> >  *         *b4_*zip* * mpr ;
> >
> > p0_zip = *1* / (*1* + exp(-eta_zip));
> >
> > eta_nb = b0_nb + b1_nb * charlson + b2_nb * no_of_diag +
> >
> >  *        b4*_nb * *mpr* +b5_nb * gap;
> >
> > mean = exp(eta_nb);
> >
> > p0 = p0_*zip* *+
> > (**1*-p0_zip)*exp(-(no_of_ed+(*1*/k))*log(*1*+k*mean));
> >
> > p_else = (*1*-p0_zip)*
> >
> > exp(lgamma(no_of_ed+(*1*/k)) - lgamma(no_of_ed+*1*) - lgamma(*1*/k) +
> >
> > no_of_ed*log(k*mean) - (no_of_ed+(*1*/k))*log(*1*+k*mean));
> >
> > if no_of_*ed*=*0* then loglike = log(p0);
> >
> > else loglike = log(p_else);
> >
> > model no_*of*_ed ~ general(loglike);
> > *
> >
> > run*;
> >
> > **
> >
> > *I got a error in SAS log:*
> >
> > NOTE: Execution error for observation 21763.
> >
> > NOTE: PROCEDURE NLMIXED used (Total process time):
> >
> > I do not know what's wroung, Is the code wrouge or something wroung
> > in the
> > data.
> >
> > thank you
> >
> > baogong
> >
>
> Baogong,
>
> Why all the asterisks (*) in your e-mail?  It makes things a bit
> difficult to read.  I don't see a consistent pattern to the
> asterisks which would indicate that these asterisks were inserted
> by the program that you use to read and send e-mail.
>
> Your problem is likely due to the default initialization of
> parameters in NLMIXED along with some large values for one or
> more of the predictor variables.  NLMIXED initializes the
> parameters to 1 by default.  If you have large values of some
> predictors, then you could be computing eta_nb too large to be
> able to compute exp(eta_nb).  Try initializing your parameters
> to zero using
>
>    parms b0_zip b1_zip b2_zip b4_zip
>          b0_nb b1_nb b2_nb b4_nb b5_nb 0;
>
> HTH,
>
> Dale
>
>
> ---------------------------------------
> Dale McLerran
> Fred Hutchinson Cancer Research Center
> mailto: dmclerra@NO_SPAMfhcrc.org
> Ph:  (206) 667-2926
> Fax: (206) 667-5977
> ---------------------------------------
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam protection around
> http://mail.yahoo.com
>



--
Baoogng Jiang
Department of Agronomy
Lousisana State University