Date:         Mon, 15 Jun 1998 08:17:39 -0700
Reply-To:     Andrew Dabalen <dabalen@ARE.BERKELEY.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Andrew Dabalen <dabalen@ARE.BERKELEY.EDU>
Subject:      Re: ERROR: Floating Point Overflow
Comments: To: bamford@WECKER.COM
In-Reply-To:  <v0310280cb1a755e54e9e@[199.4.81.107]> from Lynn Bamford at "Jun
              12, 98 03:15:59 pm"
Content-Type: text/plain; charset=US-ASCII

Floating point problems usually occur when you are doing an illegal
operation: like dividing a number by zero, etc.
 
 
 
 
 
> Hi All!
>
> I'm running a probit model (see below for exact coding)  in SAS 6.12.  I'm
> encountering an error that I've never seen before - "Floating Point
> Overflow".  Could anyone tell me what this means and what would be some
> reasonable solutions to the problem.  I've tried fooling around with the
> convergence constraint and nothing happens - i.e. error remains.  I also
> checked the distribution of my dependent variable (IMED) :
> 18,452 observations = 0 and 402 observations = 1.
>
> Thanks for your help.
>
> *****************************************************************
>
>
> proc logistic descending outest=betas;
> model imed=%varsa() /link=probit converge=.001 ;
> run;
>
> NOTE: PROC LOGISTIC is modeling the probability that IMED=1.
> ERROR: Floating Point Overflow.^MERROR: Floating Point Overflow.^MERROR:
> Floating Point Overflow.
>
> *****************************************************************
>
> --
> Lynn Bamford
> bamford@wecker.com
> 415-898-2255 phone
> 415-898-2260 fax
>
 
======================================================================
Andrew Dabalen                      University of California
Ph:510-643-5418                     Agricultural and Resource Economics
Fax:510-643-8911                    330 Giannini Hall
Email:dabalen@are.berkeley.edu      Berkeley, CA 94720
=======================================================================