Date:         Wed, 13 Feb 2008 11:47:39 -0600
Reply-To:     Yu Zhang <zhangyu05@GMAIL.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Yu Zhang <zhangyu05@GMAIL.COM>
Subject:      Re: Finding X's that are 10-day apart
Comments: To: "Jue A. Wang" <wang@stat.rutgers.edu>
In-Reply-To:  <1202920564.47b31c7447b51@outlier.rutgers.edu>
Content-Type: text/plain; charset=ISO-8859-1

Hi,

For any observation which dosen't has a  ealier 10 days counterpart, you
will choose a close one for your calcualtion. if I understood this
correctly, here is the code to get you started.

HTH

Yu
*

DATA* S;

INPUT _DATE :yymmdd10. X;

CARDS;

2007-04-29 35.4375

2007-04-24 35

2007-04-21 34.5

2007-04-20 34.25

2007-04-17 33.5

2007-04-12 33.5

2007-04-11 32.75

2007-04-10 32

2007-04-09 31.8125

2007-04-06 31.82

2007-04-05 31.875

2007-04-02 31.875

2007-04-01 32

;
*

RUN*;

title1;

footnote1;
*

proc* *sql*;

select a._date format=mmddyy10.,a.x ,b._date as date format=mmddyy10.,b.x asx1,

a._date-b._date as diff from s as a, s as b

where a._date>=b._date and *0*<=a._date-b._date<=*10
*

group by a._date

having calculated diff=max(diff);
*

quit*;







On Feb 13, 2008 10:36 AM, Jue A. Wang <wang@stat.rutgers.edu> wrote:

> My data contains a sample of X observed during certain period of time
> (DATE).
> Part of the data is enumerated below.
>
>
> DATA S;
> INPUT _DATE $1-10 X;
> CARDS;
> 2007-04-29              35.4375
> 2007-04-24                   35
> 2007-04-21                 34.5
> 2007-04-20                34.25
> 2007-04-17                 33.5
> 2007-04-12                 33.5
> 2007-04-11                32.75
> 2007-04-10                   32
> 2007-04-09              31.8125
> 2007-04-06                31.82
> 2007-04-05               31.875
> 2007-04-02               31.875
> 2007-04-01                   32
> ;
> RUN;
>
>
>
> I need to create a new variable Y which is the difference between two X's
> that
> are recorded (roughly) 10 days apart,  for example, the first X is
> recorded on
> 2007-4-29, so the first Y should be calculated as the X on this date minus
> the
> X on 2007-4-19
>
> =35.4375-34.25
>
> the second X is on 2007-4-24, since there is no record 10 day earlier on
> 2007-
> 4-14, I'd use the X closer to 2007-4-14 which is on 2007-4-12, so the
> second X
> =35-33.5
>
> I need your help that
>
> (1)If the data is complete that each record has a corresponding 10-day
> earlier
> record, but the distribution of DATE is not uniform(there is no pattern
> between adjacent DATE's) , then how can I create Y?
> (2)If the data is incomplete, in that not every X has a corresponding
> 10-day
> earlier X, how can I create Y?
>
>
> Thank you
>
> Andrew
>
>
>
> -------------------------------------------------
> This mail sent through IMP: http://horde.org/imp/
>