Date:         Wed, 13 Oct 1999 17:54:05 GMT
Reply-To:     Kattamuri.Sarma@RESPONSEINSURANCE.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Kattamuri.Sarma@RESPONSEINSURANCE.COM
Subject:      Re: Constrained Max.Likelihood in SAS.
Comments: To: "Tirthankar C.P" <tir@IGIDR.AC.IN>
Content-type: text/plain; charset=us-ascii

Depends on what sort of constraints. If you are estimating a multi equation
system, in which you
   are imposing cross equation constraints, yes, you can write your system
explicitly and use
   FIML in SAS.  I think PROC MODEL. is where to look.
     Suppose you are estimating two investment equations one for company 1,
and another for company 2.
               you equations will be   I1 = a1 + b networth.
                                                           I2 = a 2  + b
networth
               I have explicitly written my equations such that the
coefficient on networth is b in both cases.
              This is what I mean by writing the constraints explicitly
into the equations.
               Having done that I think there is a command called
estimate....
    Similarly in thecase of  nonlinear estimation,,
     I suggest you look into SAS/ETS.   My ETS manuals are not with me at
this moment, otherwise I
    could have given you the syntax..
          I think there is an example of estimating the Cobb-Douglas
production function where
          the paramers on two inputs are constrained to equal to 1. If one
coefficient is alfa the other
          is 1- alfa ..   And you give initial values.
    But, I am not sure how you can put constraints in an ARIMA model. ARIMA
models are in general
    unconstrained.

    I hope this will give an initial start.
    best of luck
    Kattamuri Sarma