Date:         Mon, 8 Jan 2007 06:21:13 -0800
Reply-To:     saikumarkorrapati@GMAIL.COM
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         saikumarkorrapati@GMAIL.COM
Organization: http://groups.google.com
Subject:      Re: Excel Solver concept in sas
Comments: To: sas-l@uga.edu
In-Reply-To:  <BAY103-F15BE95E3FCCB983D1071B9B0BC0@phx.gbl>
Content-Type: text/plain; charset="us-ascii"

Hi,

My understanding for the proeblem is its a linear programming problem (
David : excuse me if am wrong) .

hema, u r trying to maximise the correlations subject to the
constraints are given weights. so here u already calculated the
correlations individually for the constranits and i think
recomputations is not needed and only optimium solution is needed.

this can be done by proc lp  where u hav lot of options even it is
infeisible

u have to secify the no of interations and the optimum solution will be
provided.


i dared to given this suggestion infront of SAS gurus as i done my
masters in operations research ( OR) with SAS certificaton for base
programming

thanks
sai kumar

David L Cassell wrote:
> hema_dave15@YAHOO.COM wrote back:
> >
> >
> >David L Cassell <davidlcassell@MSN.COM> wrote: hema_dave15@YAHOO.COM wrote:
> > >
> > >Hi,
> > >
> > >   Can anyone tell me how can i do the same manipulation in SAS like the
> > >Solver does in Excel ?
> > >
> > >   Regards,
> > >   Hema
> >
> >I assume that you mean the Solver Add-in, and not something else.
> >
> >The Excel Solver Add-in has two parts, and I don't know which
> >one you're after.  There is a linear/integer programming component
> >which uses standard simplex methods with branch-and-bound
> >solution methods.  PROC LP does this easily, and handles MUCH
> >larger cases than Solver can cope with.  Then there is a nonlinear
> >component which uses nonlinear optimization code.  I bleieve that
> >the Solver Add-in uses Lasdon's generalized reduced gradient algorithm
> >for this.  You can use PROC NLP in these cases, and get a lot more
> >flexibility in the solution methods and capabilities.
> >
> >HTH,
> >David
> >--
> >David L. Cassell
> >mathematical statistician
> >Design Pathways
> >3115 NW Norwood Pl.
> >Corvallis OR 97330
>
> >
> >Thanks for the help but still problem unsolved :) ,..
> >I want to maximize my varaible with cretain constraints ..
> >It is more of LPP . Will Proc CPM help ?
> >I have 3 variables. I give three random weights to them and find the
> >weighted average of the 3 variables using random weights. I find the
> >correlation between the weighted average and sales variable. I want this
> >correlation to be maximum so i have to select the random weights again and
> >again with the condition that the sum of the weights is equal to one and
> >each weight is greater than zero.In Excel ,Solver(in Add-ins) works, where
> >my target variable is the correlation of Weighted average & the sales
> >variable and max option is selection and my constraints are that all
> >weights(random) are positive and they sum upto be one. So this selection
> >will give the possible random weights which will maximize my correlation ..
> >Can some one tell me how to do it in SAS?
> >thanks
> >hema
> >
> >Regards,
> >Hema.
>
> I think you may be on the wrong track here.
>
> First, tell me.  Do you have one single record of sales and your 3
> variables, or do you have many such records?  In order to get a
> 'correlation', I would assume that you would have to have a number
> of these records.
>
> If I am correct, then the proc that you ought to try is not in the
> SAS/OR module.  I suggest you try PROC REG instead.
>
> That's right, PROC REG.  Ordinary Least Squares (OLS) regression
> fits a set of parameters so that your Y (sales) has a maximal
> correlation with your linear functional of your X variables.  So I
> think oyu're making this problem way too hard.  Just regress sales
> on your three variables.
>
> Of course, you need to bear in mind that linear / non-linear
> optimizations gets derailed for pretty much the same reasons that
> regressions do, so you need to check your regression diagnostics
> too.
>
> HTH,
> David
> --
> David L. Cassell
> mathematical statistician
> Design Pathways
> 3115 NW Norwood Pl.
> Corvallis OR 97330
>
> _________________________________________________________________
> Get FREE Web site and company branded e-mail from Microsoft Office Live
> http://clk.atdmt.com/MRT/go/mcrssaub0050001411mrt/direct/01/