Date: Thu, 7 May 2009 07:10:11 -0700
Reply-To: Paul <paulvonhippel@YAHOO.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Paul <paulvonhippel@YAHOO.COM>
Subject: Re: Tobit predicted values using PROC QLIM
Content-Type: text/plain; charset=ISO-8859-1
Never mind the statistics part of the question -- I know the answer
now. The SAS part is what I really need.
On May 6, 9:45 pm, Paul <paulvonhip...@yahoo.com> wrote:
> I'm using PROC QLIM to fit a Tobit model. The Tobit model estimates
> the parameters for a normal regression of Y* on X, when all you can
> observe is a censored variable Y = max (0, Y*).
> The procedure can output predicted values for Y, but these predicted
> values are not the conditional expectation for Y -- i.e., not E(Y|X).
> Instead, the predicted values are a censored version of the
> conditional expectation for Y* -- i.e., max (0, E(Y*|X)).
> A SAS question: Does anyone know how to get SAS to produce E(Y|X)? I
> could hand-code it, sure, but I wonder if QLIM or another PROC will
> just give it to me.
> A statistics question: Does anyone know how much of a difference there
> is between the result I'm getting and the result I want. I.e., how
> well does max (0, E(Y*|X)) approximate E(Y|X)?