Date:         Mon, 6 Nov 2000 18:57:28 +0100
Reply-To:     Rodolphe Thiebaut <rodolphe.thiebaut@DIM.U-BORDEAUX2.FR>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Rodolphe Thiebaut <rodolphe.thiebaut@DIM.U-BORDEAUX2.FR>
Organization: Universite Bordeaux I
Subject:      Re: MLE

OK. I begin to ask a question.
I have written a code with IML to perform a mixed model as with Proc MIXED
using call nlpnra. When I run it, the algorithm converge (with true
estimates) but I have the message "NOTE : I/O required temporary file to be
opened" repeated x times. Note that I calculate matrix for fixed and random
effects for every patients as I have missing data.

Is there anybody konwing what is this message and if it exists an index of
error's message using IML ?

Rodolphe.

<steven_wang@my-deja.com> a écrit dans le message news:
8u6n2k$fd0$1@nnrp1.deja.com...
> Hi:
>
> Thank you for your response. I also got two other emails regarding MLE
> in SAS. See below. Hope this will bring some more discussions on this
> topic. I'll also do some testing runs and report the results.
>
>
> Cheers,
>
> Steven
>
>
> 1.
> >As far as I know, SAS does not have such a procedure.  If you can
> recast the
> >likelihood function as a linear or non-linear regression, a mixed
> model, or
> >an optimization problem, SAS has several procedures to perform the
> >equivalent of maximum likelihood estimation:  PROC REG, PROC NLIN, PROC
> >MIXED, or PROC IML.
>
>
> 2.
> >    Try Proc NLP in SAS/OR. Here is what I did with a similar problem.
> lnL
> >refers to the equation for the log of the likelihood function:
> >
> >
> >proc nlp data=XXX perror;
> >  parms a=x, b=x, c=x;     /* set starting values for parameters */
> >  bounds b <= 1      ;     /* define constraints on parms if you have
> any */
> >  max lnL;
> >  profile a b c / alpha=0.95;      /* if you want 95 % LPCIs for parms
> */
> >  lnl=0;
> >  do .... ;      /* loops over x and y */
> >    lnL= ...     /* your logL function */
> >  end;
> >
> >
> >   I hope this helps. Please let me know what other suggestions you
> get, as I
> >am always looking for alternative ways.
>
>
> In article <8u5r09$rob$1@news.u-bordeaux.fr>,
>   "Rodolphe Thiebaut" <rodolphe.thiebaut@dim.u-bordeaux2.fr> wrote:
> > Use proc IML with nonlinear optimization functions.
> >
> > Rodolphe.
> >
> > <steven_wang@my-deja.com> a écrit dans le message news:
> > 8tut5q$kbl$1@nnrp1.deja.com...
> > > Hi guys:
> > >
> > > I am new here. I want to know what is the SAS Proc to use for
> maximum
> > > likelihood estimation. For example, I have a likelihood function:
> > >
> > > L=a+b*y+c*x+e
> > >
> > > How can I get the MLE for the parameters using a SAS Proc?
> > >
> > > Thanks!
> > >
> > > Steven
> > >
> > >
> > > Sent via Deja.com http://www.deja.com/
> > > Before you buy.
> >
> >
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.