|
On Tue, 29 Jun 2004 15:46:02 -0500, Dunn, Toby <tdunn@TEA.STATE.TX.US>
wrote:
>Brian,
>
>The problem with your solution is that it requires knowing how many
>missing values in a row you have before you create your lags. Thus with
>out a lead, interleaving a data set upon itself, or a DoW loop you
>wouldn't know.
>
>Toby Dunn
>
>-----Original Message-----
>From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
>Brian Shilling
>Sent: Tuesday, June 29, 2004 3:44 PM
>To: SAS-L@LISTSERV.UGA.EDU
>Subject: FW: Lag Question
>
>What about increasing the lag using an array, or multiple lag functions?
>
>LAG assumes LAG1 - one lag back. You could try z = lag(x) zz = lab2(x)
>zzz =
>lag3(x) and test the missing values that way.
Hi, Toby,
Good question.... IMHO, Brian meant something like below -- which is
quite...hmmm... NOT pretty. :-)
Cheers,
Chang
data one;
input x $ @@;
cards;
1 2 . 4 5
1 . . 4 5
1 . . . 5
. . . 4 5
1 2 3 . .
;
run;
%macro ninetyNine;
%local i;
data two(drop=lag1_x--lag99_x);
set one;
%do i = 1 %to 99;
lag&i._x = lag&i.(x);
%end;
new_x = x;
if missing(x) then do;
if not missing(lag1_x) then new_x = lag1_x;
%do i = 2 %to 99;
else if not missing(lag&i._x) then new_x = lag&i._x;
%end;
end;
run;
%mend;
%ninetyNine
proc print data=two;
var x new_x;
run;
/* on log
Obs x new_x
1 1 1
2 2 2
3 2
4 4 4
5 5 5
6 1 1
7 1
8 1
9 4 4
10 5 5
11 1 1
12 1
13 1
14 1
15 5 5
16 5
17 5
18 5
19 4 4
20 5 5
21 1 1
22 2 2
23 3 3
24 3
25 3
*/
| 
