Date:         Sun, 18 Mar 2001 17:01:29 GMT
Reply-To:     Roger Lustig <julierog@IX.NETCOM.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Roger Lustig <julierog@IX.NETCOM.COM>
Subject:      Re: Generate the third variable based on the categories of first
Content-Type: text/plain; charset=us-ascii

wei cheng wrote:

> Hi SAS-Lers,

> Sorry for the late Friday question.

No problem.  We're open 24/7.

> If I have two categorical variables,
> Eg. variable A has levels x, y, z
>     variable B has levels a,b,c,d

> and I need to create a third variable C based on them like:

> A        B        C
> y        b        a2b2
> x        b        a1b2
> x        d        a1b4
> z        a        a3b1
> y        c        a2b3
> .......

> We don't know the levels of A and B in advance, but we need to have variable
> C to have the levels of A and B.

Ya Huang already suggested a way of using FREQ and FORMAT to get the
order statistics and then concatenate the formatted values.  There are
other implementations of the same approach, such as doing a FREQ on each
variable individually, an output data set for each, each of those data
sets becoming the CNTLIN data set to create a format.  Then create your
new variable by concatenating formatted values of each input variable.
(This should be very easy to program into a macro, by the way.)

If you just need a unique value for C, one that does not reflect the
order statistics of A and B, you can simply concatenate the two
variables (if they're character) or use a transformation like

2**A * 3**B

if they're natural numbers.

Roger