Date: Mon, 8 Apr 1996 19:11:05 GMT
Reply-To: Sean Sullivan <sps_ba@IX.NETCOM.COM>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: Sean Sullivan <sps_ba@IX.NETCOM.COM>
Organization: Netcom
Subject: Re: more about OBS and _N_
Chun Huang <CHUN@MANA.MEDSURF.COM> wrote:
Simply put, _N_ counts the number of interations of the datastep, not
the number of observations in the data.
>Dear SAS-Lers:
>
>Yesterday Sudhakar asked how to get the sequence number of observation
>in a sas data set and use it in the analysis. In my response to the
>message, I suggested that a sum statement (e.g., OBS_NUM+1) can be
>used in this situation. In response to Sudhakar's message, Whittington
>et al. suggest using the automatic variable, _N_. Several people also
>sent messages to me and kindly reminded me that using _N_ is the best
>way to do this. However, I think it may or may not be the best way.
>Joe has pointed out the potential problems with _N_. Here I
>would like to provide another very simple example:
>
>DATA TEST;
>INPUT ID X;
>IF ID=2 THEN DELETE;
>IF ID NE 3;
>NUM_OBS+1;
>N=_N_;
>CARDS;
>1 11
>2 12
>3 13
>4 14
>5 15
>;
>RUN;
>PROC PRINT;
>RUN;
>
>The SAS System 1
>08:58 Thursday, April 4, 1996
>
>OBS ID X NUM_OBS N
>
>1 1 11 1 1
>2 4 14 2 4
>3 5 15 3 5
>
>This example shows that the value of _N_ also increases by 1 when a
>IF---THEN DELETE statement is executed or when a subsetting IF is
>executed with a false condition; but the sum statement does not (as
>long as it used after the IF statements). _N_ may give us
>unexpected and misleading value in this situation.
>
>Chun Huang, Ph.D
>
>
>Chun Huang, Ph.D.
>Hawaii Osteoporosis Center
>Email: chuang@mana.medsurf.com
***************************************
Sean P. Sullivan
MICS Administrator
Key Services Corporation
NY-31-22-0303
22 Corp Woods Blvd
Albany, NY 12211 USA
1-518-391-6009
s_sullivan@keycorp.e-mail.com
sps_ba@ix.netcom.com
******************************************
| 
