Date:         Tue, 1 Oct 1996 20:57:30 +0200
Reply-To:     Elik Jucha <jucha@NETVISION.NET.IL>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Elik Jucha <jucha@NETVISION.NET.IL>
Subject:      Treating unbalanced data set?

Dear Clare!
 
I suggest you to use PROC MIXED instead PROC REG. Proc Mixed
treats unbalanced data and does regression analysis.
 
Elik Jucha
Statistix - Statistics & Data Analysis
Tel/Fax : 972-9-461604
Hod-Hasharon, Israel.
 
-------------- Original posting
----------------------------------------------------------------------------
----------------
Does anyone know if there is a way to automatically treat unbalanced data
without having to drop problematic observations one by one?
 
For example when I use proc reg with an unbalanced data set I get a "not
fully ranked" message for certain observations.  I then end up deleting
those observation manually.  Often I am running several models in the
same batch with different variables and thus I have to run each model
separately dropping different observations.  Is there a more efficient
way to do this?
 
I am estimating the regression using fixed effect.
 
Thanks
---
E. Jucha