Date:         Tue, 2 Aug 2011 19:59:56 -0500
Reply-To:     "Data _null_;" <iebupdte@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "Data _null_;" <iebupdte@GMAIL.COM>
Subject:      Re: Progressive calculation of a variable using its own lagged
              values
Comments: To: Monica <WBS.PhD@gmail.com>
In-Reply-To:  <201108030004.p72AqFMN015978@wasabi.cc.uga.edu>
Content-Type: text/plain; charset=ISO-8859-1

RETAIN EWMA as in this example.  Please check formula carefully.

data ewma;
   input Date:mmddyy. Ret Std Lambda;
   if _n_ eq 1
      then ewma = std;
      else ewma = sqrt(lambda*ewma**2 + ewma + (1-lambda)*ret**2);
   retain ewma;
   format date mmddyy.;
   cards;
02/01/00    0.12  0.1          0.94
03/01/00    0.11  0.2          0.94
04/01/00    0.04  0.3          0.94
05/01/00    0.10  0.4          0.94
06/01/00    0.05  0.5          0.94
07/01/00    0.15  0.6          0.94
;;;;
   run;
proc print;
   run;
On Tue, Aug 2, 2011 at 7:04 PM, Monica <WBS.PhD@gmail.com> wrote:
> Dear all,
> I have an apparently simple question which I cannot sort out by myself.
> Let's say I have the following dataset:
> Date (t)    Ret   Std   Ewma Lambda
> 02/01/00    0.12  0.1          0.94
> 03/01/00    0.11  0.2          0.94
> 04/01/00    0.04  0.3          0.94
> 05/01/00    0.10  0.4          0.94
> 06/01/00    0.05  0.5          0.94
> 07/01/00    0.15  0.6          0.94
>
>
> I need to construct the Ewma variable in the following way:
> for first observation (t=02/01/00) Ewma(t)=Std(t)=0.1
> for all other obsevations t:
> Emwa(t)=sqrt(Lambda*Emwa(t-1)*Emwa(t-1)+(1-Lambda)*Ret(t)*Ret(t))
> Where Emwa(t-1) is simply the past value of Emwa.
> Could you suggest an efficient way to do it?
> Many thanks for your help,
> Monica
>