|
Jim Savage a icrit dans le message <37202bab.22419191@news.earthlink.net>...
>I am trying to do a simple calculation in which I want to have the values
of a
>new variable build up based on the values of an input variable and then
decay
>when the input variable declines. This is being used in Advertising
research
>where we calculate a decayed/smoothed measure based on TV Gross Rating
Points
>which are usually spikes or pulses. Here's the logic that I am trying to
do:
>
>Decayed GRP's for week 1 = Input GRP's for week1
>Week 2 decayed GRP's = Input GRP's for week 2 + 80% of the Wk1 decayed
GRP's
>Week 3 decayed GRP's = Input GRP's for week 3 + 80% of the Wk2 decayed
GRP's
>
>After these decay operations are done for each week the result is raised to
the
>.65 power.
>
>This works really well in a spreadsheet, but I thought it would be easier
in
>SAS, since that's where I'm going to do the model anyway.
>
>Here's what I tried with no success. I think there is something I don't
>understand about the LAG function.
>
>Data two; set one;
> by market;
>retain DECAYGRP;
>DECAYGRP=GRP;
>DECAYGRP=lag1(DECAYGRP)*.8);
>DECAYGRP=SUM(of GRP DECAYGRP););
>if first.market then DECAYGRP=GRP; /* want to start over for each market
*/
>run;
>
>The first 2 observations are calculated correctly, but from then
>on the values are wrong. Anyone have any thoughts as to
>what I can do to fix this problem?
>
>Jim Savage at ljas@earthlink.net
>
>Savage Genealogy: http://home.earthlink.net/~ljas/savage/index.htm
>Landers Genealogy: http://home.earthlink.net/~ljas/index.htm
>
>
Hi ,
I don't know if i understood exactly your problem but
try replacing thess lines
>retain DECAYGRP;
>DECAYGRP=GRP;
>DECAYGRP=lag1(DECAYGRP)*.8);
>DECAYGRP=SUM(of GRP DECAYGRP););
>if first.market then DECAYGRP=GRP;
by these
if first.market then DECAYGRP=GRP;
else
DECAYGRP=lag(DECAYGRP)*0.8+GRP;
According to me, you don't have to use the retain statement
as its sounds to perform the same work as lag function
I don't really know the function "SUM(of " that you use (I always perform
proc means);
Good luck,
Regards,
Fabien
| 
