Thanks Ian,

I think some variation on the second suggestion might work. Here is simply what I need for each variable:- if OLD IN (.) then OLD_I=1; else OLD_I =0;

The final data set should contain the original variable names + new variable names (original name plus some extension)as I need to match the old and new variable names in an imputation program as with syntax: old (old_I).

Sorry for not being clearer- it took me a while to figure out exactly what I needed.

Mah-J

-----Original Message----- From: iw1junk@comcast.net [mailto:iw1junk@comcast.net] Sent: Friday, December 02, 2005 1:39 PM To: SAS(r) Discussion Cc: mah-j@statworks.com Subject: Re: array help

Mah-j, You cannot hold status of 55 variables in one variable with the same record structure ( not even with two variables). You could have

array v (55) ... ; array stat (55) ; do i = 1 to dim ( v ) ; /* I am not clear on what is bad */ stat[i] = (v[i] not in ( . )) ; end ;

or perhaps a better structure would be

data problems ( keep = obs name value ) ; set ... ; array v (55) ... ; obs + 1 ; /* or better keys to record */ do i = 1 to dim ( v ) ; stat = (v[i] ^= .) ; if stat = 0 then do ; name = vname ( v[i] ) ; value = v[i] ; output ; end ; run ;

It all depends on what you plan to do with the information. For example

proc freq data = problems ; table obs ; /* how many problems in each observation */ table name ; /* how many problem in each variable */ table value ; /* summary count of bad values */ run ;

Ian Whitlock ================= Date: Fri, 2 Dec 2005 11:19:27 -0500 Reply-To: mah-j@statworks.com Sender: "SAS(r) Discussion" From: "mah-j@statworks.com" <msoobader@VERIZON.NET> Subject: Re: array help Comments: To: Jim Groeneveld <jim1stat@YAHOO.CO.UK> In-Reply-To: <200512021556.jB2EsCN7027273@mailgw.cc.uga.edu> Content-type: text/plain; charset=iso-8859-1 Thanks Jim, This is my dilemma. I have around 55 variables. I'm not sure how to get around ^=. - the problem is some variables have numbers 0,1 some 1,2,3,4,5 and some 0-1000- so all these need to be included with .L (legitimate skip) and .N (not in universe) and not imputed. Maybe the other side of this will work. If var(i)=.P or var(i)=.M or var(i)=. then x=0;? My previous email - noted however that there was a problem with my method as you point out -x always get set to 1. Still I'm not sure you can just use one variable to do this -I think maybe the way to approach it is to duplicate the entire dataset so you have 2 variables ( var1 + var1x)- and then var1x identifies the imputation condition for var1.

Thanks Mah-J