Date: Tue, 14 Oct 1997 04:11:16 +0100
Reply-To: Jeroen Liefhebber <datastep@PI.NET>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Jeroen Liefhebber <datastep@PI.NET>
Organization: Datastep
Subject: Re: A question on PROBIT
Content-Type: text/plain; charset=us-ascii
Yize Wang wrote:
>
> Hi, SAS-Lers:
>
> I am using the probit model to see how personal characteristics influence
> the probability of getting insured. My program works like this:
>
> PROC PROBIT;
> CLASS COVERAGE;
> MODEL COVERAGE = Age Sex Occupation ....;
> RUN;
>
> I use COVERAGE=1 for those with insurance and COVERAGE=0 for those
> uninsured. What I think is that a positive coefficient tells that
> the higher this parameter, the more likely the person is insured.
>
> However, the coefficients is just opposite to what I think should
> be. I refer to SAS/STAT manual and it says that "the probability of the
> LOWER value of the dependent variable is modelled in the probit
> procedure". If this is true, what the above model tells me is how these
> independent variables contribute to the risk of being UNINSURED.
>
> I wonder any of you have experience with the PROBIT and can help me with
> this. Thanks in advance.
>
> --
> Yize Wang
> Department of Health Care Systems
> University of Pennsylvania
Try
PROC PROBIT DATA=.. ORDER=INTERNAL ;
The results of probit-estimations depend on the order in which the
observations are read from the data-set you are using within each value
of the dependent class-variable.
/* Jeroen Liefhebber */
/* */
/* Datastep SAS Software Consulting */
/* Luchthavenweg 81 */
/* P.O. Box 4985 */
/* 5604 CD Eindhoven */
/* The Netherlands */
/* e-mail Jeroen.Liefhebber@datastep.nl */
| 
