Ya,

The last time I tested the data-type-independent MISSING() function I was surprised to discover that it performed noticeably faster than the condition ( LE .Z ) despite the fact that the latter seemingly had less work to do. So, I would follow Jack's advice and use MISSING() under all circumstances where all the missing values have to be embraced collectively.

Kind regards, ================== Paul M. Dorfman Jacksonville, FL ==================

>From: "Huang, Ya" <yhuang@AMYLIN.COM> > >Hi there, > >I usually use "if a=. then ..." to test if a variable's value is >missing, but was burned today, because one of the var has many different >missing values, as a result "if a=. then .." failed to detect a=.k. >I then use "if nmiss(a)=1 then ..", which works fine. I just wonder >how you guys handle this kind of situation. > >61 data _null_; >62 a=.; >63 b=(a=.); >64 put a= b=; >65 a=.k; >66 b=(a=.); >67 put a= b=; >68 a=.k; >69 b=(nmiss(a)=1); >70 put a= b=; >71 a=.k; >72 b=(a = missing); >73 put a= b=; >74 run; > >NOTE: Variable missing is uninitialized. >a=. b=1 >a=K b=0 >a=K b=1 >a=K b=0 > >I vaguely remember seeing someone use 'missing' as a key word, >but the above log shows it dose not work. > >Thanks > >Ya

_________________________________________________________________ The new MSN 8: advanced junk mail protection and 2 months FREE* http://join.msn.com/?page=features/junkmail