Date: Wed, 5 Mar 2003 17:54:08 -0800
Reply-To: Yvette <hollied@OLIN.WUSTL.EDU>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Yvette <hollied@OLIN.WUSTL.EDU>
Subject: OLS in SAS - coefficient constraints
Content-Type: text/plain; charset=ISO-8859-1
I am having some difficulty trying to figure out how to estimate their
coefficients in SAS using aggregate versus disaggregated data. My
first question is, is this possible? If possible, then does anyone
have any suggestions on dealing with this problem with regards to the
programming. For example:
I have a dependent variable (DV) at time t+1 and my independent
variables (IV) are at time t (prediction model). The issue is that
the IV are disaggregated data (that sum to consolidated totals at the
firm level) and the DV is an aggregated number (at the firm level). I
want the coefficient estimates to be consistent with a firm
cross-sectional analysis but taking into account the disaggregated
data in the estimates of IV1 and IV2.
Example: DV = intercept + IV1 + IV2 + error term
Firm ID DV IV1 IV2
1 5 7
1 3 2
1 12 3 4
2 6 3
2 9 2 7
Is it posible to program this type of equation, and if so, what might
be possible solutions? Because of data limitations I can only get the
DV on an aggregated basis but I want to examine the effects of the
disaggregated data (IV) on the predictability of DV.
I can provide additional info if necessary.
I would greatly appreciate any suggestions/comments you can offer.