Date:         Mon, 13 Nov 2006 12:13:14 -0500
Reply-To:     Richard Ristow <wrristow@mindspring.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Richard Ristow <wrristow@mindspring.com>
Subject:      Re: Question on looping syntax
Comments: cc: hillel <hilel@BGU.AC.IL>
In-Reply-To:  <45585628.7050000@bgu.ac.il>
Content-Type: text/plain; charset="us-ascii"; format=flowed

At 06:25 AM 11/13/2006, hillel wrote:

>The same calculation can be done with no need of loop.
>
>Here is the syntax for it as will as Richard Ristow do reapeat example
>.
>
>compute #sum1=sum(#sum1,num, - lag(num,6)) .
>compute sum1 = #sum1.

Right. That's the other way to do it: keep a running running sum, your
"#sum1". I think you want "lag(num,5)" instead of "lag(num,6)", though.
Remember, the running sum of 5 is of the current value through 4 back,
not through 5 back.

Your logic treats all missing values in the sequence as 0; that may or
may not be what's wanted. In particular, before the 5th case, you add
all of the previous cases; those results may be what is desired, but
could be seriously misleading.

The DO REPEAT logic could be written either way; I've written it to
treat missing values as missing.

SPSS draft output:

* ...................................................          .
LIST.
|-----------------------------|---------------------------|
|Output Created               |13-NOV-2006 12:03:21       |
|-----------------------------|---------------------------|
  ID DATUM

001    .1
002    .3
003    .5
004    .7
005    .9
006   1.1
007   1.3
008   1.5
009    .
010   1.9
011   2.1
012   2.3
013   2.5
014   2.7
015   2.9
016   3.1
017   3.3
018   3.5
019   3.7
020   3.9

Number of cases read:  20    Number of cases listed:  20


* Compute running sum of the previous 5 values, counting the   .
* current value.                                               .

NUMERIC RNINGSUM (F6.1).
COMPUTE RNINGSUM = DATUM.
DO REPEAT  LAGVAL = 1 TO 4.
.  COMPUTE RNINGSUM = RNINGSUM + LAG(DATUM,LAGVAL).
END REPEAT.

*  The same, with "running running sum" logic:                 .

NUMERIC SUM1  (F6.1).
NUMERIC #SUM1 (F6.1).

compute #sum1=sum(#sum1,DATUM, - lag(DATUM,5)) .
compute sum1 = #sum1.


LIST.
|-----------------------------|---------------------------|
|Output Created               |13-NOV-2006 12:03:22       |
|-----------------------------|---------------------------|
  ID DATUM RNINGSUM   SUM1

001    .1      .       .1
002    .3      .       .4
003    .5      .       .9
004    .7      .      1.6
005    .9     2.5     2.5
006   1.1     3.5     3.5
007   1.3     4.5     4.5
008   1.5     5.5     5.5
009    .       .      4.8
010   1.9      .      5.8
011   2.1      .      6.8
012   2.3      .      7.8
013   2.5      .      8.8
014   2.7    11.5    11.5
015   2.9    12.5    12.5
016   3.1    13.5    13.5
017   3.3    14.5    14.5
018   3.5    15.5    15.5
019   3.7    16.5    16.5
020   3.9    17.5    17.5
Number of cases read:  20    Number of cases listed:  20
++++++++++++++++++++++++++++++++++++
I promised this: Here is the test-data generator.
NEW FILE.
INPUT PROGRAM.
.  NUMERIC ID    (N3)
           /DATUM (F4.1).
.  LOOP   ID = 1 TO 20.
.     COMPUTE DATUM = (2*ID-1)/10.
.     IF (ID EQ 9) DATUM = $SYSMIS.
.     END CASE.
.  END LOOP.
END FILE.
END INPUT PROGRAM.