Date:         Tue, 10 Aug 2004 08:51:52 -0600
Reply-To:     Jack Hamilton <JackHamilton@FIRSTHEALTH.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Jack Hamilton <JackHamilton@FIRSTHEALTH.COM>
Subject:      Re: Two basic SAS datastep questions but very urgent!
Comments: To: ben.powell@cla.co.uk
Content-Type: text/plain; charset=us-ascii

Char is stored as 1-32K bytes, depending on how it is (implicitly or
explicitly) declared.  In a data step, the implicit declaration is the
length of the first string assigned to it.  A numeric is stored as 2-8
bytes (3-8 bytes on many operating systems).  with 8 the default.

Something that catches many new users by surprise:

=====
if a = 1 then
   x ='No';
else
   x = 'Yes';
=====

will give X a length of 2, because the first string assigned to it is 2
characters long.

In v6 and previous versions of SAS, the maximum char length was 200.

While a data step is actually running, numerics are kept in memory as 8
bytes, even if stored as 3 bytes, and are truncated when written out*.
This is what Paul Dorfman was referring to in a previous note.

* I think they're truncated.  If someone knows differently, please let
me know.




--
JackHamilton@FirstHealth.com
Manager, Technical Development
Metrics Department, First Health
West Sacramento, California USA

>>> "Ben Powell" <Ben.powell@cla.co.uk> 08/09/2004 11:46 PM >>>
Thanks for the tip. Is each Char stored as one byte then in SAS, with
numerics as eight bytes?

Ben.

-----Original Message-----
From: Jack Hamilton [mailto:JackHamilton@firsthealth.com]
Sent: 09 August 2004 19:17
To: ben.powell@CLA.CO.UK; SAS-L@LISTSERV.UGA.EDU
Subject: Re: [SAS-L] Two basic SAS datastep questions but very urgent!


It would probably be preferable to use a character flag set to Y/N (or
the
local equivalent).

On most platforms, a numeric variable will take 8 times more storage
space
than a one-byte character value by default (3 times in the best
case) and requires additional processing to display for output.

Using a numeric flag saves a tiny bit of typing ("if flag" instead if
"if
flag = 'Y'"), but unless you want to perform calculations on the field
(in
which case it's not really a flag), storing flags as short character
fields
is better.





--
JackHamilton@FirstHealth.com
Manager, Technical Development
Metrics Department, First Health
West Sacramento, California USA

>>> <ben.powell@CLA.CO.UK> 08/09/2004 6:32 AM >>>
To mark specific records you could use a flag, i.e set variable flag =
1 or
0 depending whether it is one of your select values or not.

For instance,

data test;
set test;
if _N_ in(7 90 100 101 103 400 523 633 635 800) then flag = 1;/*OR USE
ANOTHER METHOD TO SELECT THE RECORDS, FOR INSTANCE IF THEY ARE IN A
CERTAIN
RANGE.*/ else flag = 0; run;

data subset;
set test;
if flag = 1;
run;

or

proc freq data = test (where=(flag=1));table zipcode;run;

Incidentally, the final suggestion above will solve your second problem
and
give you a distinct list of all zipcodes should you remove the where
clause.
Alternatively you could create an output dataset:

proc freq data = test;
table zipcode / out = tf;
run;

Alternatively, try sql:

proc sql;
create table z_count as
select distinct zipcode, count(*) as qty
from test
group by zipcode;
quit;

HTH.

On Sat, 7 Aug 2004 13:19:08 -0700, Se Yan <laseyan@GMAIL.COM> wrote:

>1.  Is there any method in SAS that can mark some specific records in

>my dataset?  I have a dataset with 10000 records, but i have ten
>records that are special and need to be marked, so that in the
further
>procedures I can choose to include or exclude these records, but i
>don't want to split the dataset into two files.  How can I do that?
>
>2. I have 10000 individual records, but there is a variable
"zipcode",
>which only has 500 different values, i.e, average 20 persons per
>zipcode in the dataset.  Which command can give me the list of these
>500 different values of zipcodes?
>
>I know the problems are quite basic.  BUt i madly need help!