Date:         Wed, 23 Jun 2004 13:53:47 -0400
Reply-To:     Karen Olson <karen.olson@TCH.HARVARD.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Karen Olson <karen.olson@TCH.HARVARD.EDU>
Subject:      PROC EXPAND / time series recs when dates are missing

I used proc freq to count records by date and saved the frequencies in a
dataset with an ODS OUTPUT statement.  Occasionally there are dates with
no events and so the output dataset does not create a record for that
date.  However, I want to do some time series analyses and need a record
for every date.  Therefore, I need to create records for the missing dates
where the frequency of the event = 0.

I thought maybe PROC EXPAND would do this.  I can get new records with the
missing dates, but the value of the frequency variable ends up being some
interpolated value.  Is there a way to use this procedure and get the
value to always be zero if the date had been missing in the first place?

For example, I start with this:

hosp        vdate    Frequency

   2    05/25/2004       167
   2    05/26/2004        53
   2    05/31/2004       243
   2    06/01/2004       213
   2    06/02/2004       229

And I want this:

hosp       vdate    Frequency

   2    25MAY2004     167.000
   2    26MAY2004      53.000
   2    27MAY2004       0.000
   2    28MAY2004       0.000
   2    29MAY2004       0.000
   2    30MAY2004       0.000
   2    31MAY2004     243.000
   2    01JUN2004     213.000
   2    02JUN2004     229.000

Thanks,

Karen Olson
Children's Hospital Boston
karen.olson@tch.harvard.edu