Date:         Fri, 22 Sep 2006 16:03:43 -0400
Reply-To:     Zach Peery <zpeery@NATURE.BERKELEY.EDU>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Zach Peery <zpeery@NATURE.BERKELEY.EDU>
Subject:      Problem with PROC FREQ

Hi All,

I have a dataset setup to try and estimate the frequency of occurrence of
variables states for all possible PAIRWISE comparisons of variables.
(e.g., State I and J of Variable 1 vs State P and R of Variable 2).  EZ,
except that I need to include combinations of States in the output dataset
for which there are zero occurrences.  In the example input dataset at the
bottom there were no observations with State J for variable 1 and State Q
for variable 2 so the output dataset should have a row that looks like
this:

VarA   VarB   StateOfVarA    StateOfVarB   COUNT
var1   var2       J             Q            0

I've tried:

proc freq data=data noprint;
tables  VarA*VarB*StateOfVarA*StateOfVarB / out=Frequencies sparse;

but the problem with this approach is that it outputs states that dont
exist (e.g. State Z for Variable 1).

Example input data set:
VarA   VarB   StateOfVarA    StateOfVarB
var1   var2       I             P
var1   var2       I             Q
var1   var2       J             P
var1   var3       I             Y
var1   var3       I             Z
var1   var3       J             Y
var1   var3       J             Z
var2   var3           etc...
etc....


In reality there are many variables and many states per variable...Many
thanks for any help.  Zach