Date:         Wed, 21 Aug 1996 07:54:52 -0700
Reply-To:     Karsten Self <kmself@IX.NETCOM.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Karsten Self <kmself@IX.NETCOM.COM>
Subject:      Re: Using a variable's values to create SAS source code
Comments: To:
          /G=Robert/S=Agostinelli/O=astra_usa_inc/OU1=ASTRA1@MHS-ASTRAUSA.ATTMAIL.COM

If I'm reading you right, no macro required, though possibly a different
data structure.  Suppose you 'condition' to 'HiAge' and 'LoAge'.  You can
use these values in comparison operators.  Four cases to consider:
 
hi= missing, lo= missing:         don't perform comparison
hi= non-missing, lo= missing:     test age < hi
hi= missing, lo= non-missing:     test age > lo
hi= non-missing, lo= non-missing: test lo < age < hi
 
You indicate that other types of comparison are possible, eg:  sex.  Add a
field and perform similar logic.
 
There are two methods for generating code based on data values.  One uses
macros and the call symput statement, eg:
 
 data _null_;
    mname= 'fred'; x= 1; y= 2;
    call symput( mname, put( x, 8.0 ));
    call symput( 'sue', put( y, 8.0 ));
    run;
 
  ...creates two macrovariables 'fred' (value of 'mname') and 'sue' with
values '1' and '2' respectively.  For your application, the macrovariable
overhead is probably too great -- macrovariables use memory.
 
An alternate method is the call execute statement.  This resolves its
arguments to generate code which is executed following the close of the data
step.  I use this technique, though in your case it will not generate code
to be applied in the current data step.  Again, you would be producing a
program with a block of statements corresponding to each observation in your
data.
 
At 02:36 PM 8/20/96 -0500, you wrote:
>     Hi folks,
>
>     I have a SAS dataset ( with 1000's of observations ) that has 4
>     variables:
>
>
>     Patient  Age  Sex   Condition
>     -------  ---  ---   ------------------------------------------
>
>       1       21   M    If age > 12 and age < 26
>       2       63   F    If age < 18
>       3       45   M    If age > 18 and sex = M
>       .
>       .
>       .
>     etc.
>
>
>     The condition variable is different for each observation. I have
>     already written the SAS code that parses the condition into three
>     variables: a low age range, a high age range, and the sex condition.
>     In reality, the condition also contains criteria for race and other
>     parameters as well.
>
>     However, I am looking for a cleaner solution than having to parse the
>     text. I would like to be able to use the value of the condition
>     variable as part of the SAS program that evaluates the observation as
>     pass/fail. To do this, I know that macros are required to build
>     dynamic SAS code.
>
>     Therefore, I want to create a 5th variable that flags whether or not
>     the current obs. meets it's own condition.
>
>     For example, patient 1 would pass it's own condition, and the flag
>     would be set to 'pass'. Patient 2 would fail since the patient is 63
>     years old and the condition to pass requires them to be less than 18.
>     Patient 3 would pass.
>
>     Does anyone know what the SAS macro code should be to perform this
>     task?
>
>     Thanks for your help.
>
>
----------------------------------------
Karsten Self / kmself@ix.netcom.com
 
What part of gestalt don't you understand?