topic Re: Create tabulation of two unique values (like times-table of old) in SAS Procedures

Create tabulation of two unique values (like times-table of old)

alawton — Mon, 13 Jan 2014 18:45:30 GMT

I was asked to create a grid of the combinations of any two diagnoses by unique claim (the grid would be set up like those old multiplication tables from elementary school). However the diagnoses exist in separate variables for primary diagnosis and secondary diagnosis; each observation is a unique claim. I want to identify the combinations so that combination A/B would be counted as the same as B/A. A simple proc tabulate would did not work because A/B and B/A were counted separately. I came up with a round-about way of doing it, but I feel there must be a more straight-forward approach. Can you help me get there? My longer code is:

proc sql;

create table claimsub as

select claims2.*,

t.Diagnosis as Diagnosis1,

t2.Diagnosis as Diagnosis2

from claims2

inner join tops2 t on claims2.diag1=t.diag1

inner join tops2 t2 on claims2.diag2=t2.diag1

union

select claims2.*,

t2.Diagnosis as Diagnosis1,

t.Diagnosis as Diagnosis2

from claims2

inner join tops2 t on claims2.diag1=t.diag1

inner join tops2 t2 on claims2.diag2=t2.diag1;

quit;

proc sort data=claimsub;
by Diagnosis1 Diagnosis2;
run;

proc sort nodupkey data=claimsub out=claimsub2;
by claim;
run;

proc tabulate data=claimsub;
class Diagnosis2 Diagnosis1;
table Diagnosis2, Diagnosis1 * N = '';
run;

Thanks!

Re: Create tabulation of two unique values (like times-table of old)

Reeza — Mon, 13 Jan 2014 19:06:43 GMT

I may be missing something because this does look like a basic proc freq.

Can you post an example of what your data looks like?

Re: Create tabulation of two unique values (like times-table of old)

Fugue — Mon, 13 Jan 2014 19:12:51 GMT

If I understand your data structure correctly (one row per claim, with two diagnostic variables), you can simply create two new variables and assign the "smallest" code to variable 1 and the "largest" code to variable 2.

Further, you could do all the processing in one PROC SQL step (no need to the proc sort or the proc tabulate) using GROUP BY and ORDER BY clauses.

Re: Create tabulation of two unique values (like times-table of old)

Tom — Mon, 13 Jan 2014 19:14:48 GMT

Not sure I understand the problem, but if you want to make a triangular result matrix from PROC FREQ then sort the values in the two variables.

data want ;

set have ;

if diag1 > diag2 then do; swap=diag1; diag1=diag2; diag2=swap; end;

drop swap;

run;

proc freq ;

tables diag1*diag2 ;

run;

Re: Create tabulation of two unique values (like times-table of old)

alawton — Mon, 13 Jan 2014 19:31:22 GMT

This answer works because my variables are numeric and solves my dilemma here. But taking the thought a step further, what should I do if my variables were character instead numeric (diag1desc/diag2desc instead of diag1/diag2)?

claim	diag1	diag1desc	diag2	diag2desc	Diagnosis1	Diagnosis2
214991610	30981	PROLONG POSTTRAUM STRESS	30981	PROLONG POSTTRAUM STRESS	06 - 30981 - PROLONG POSTTRAUM STRESS	06 - 30981 - PROLONG POSTTRAUM STRESS
216531478	311	DEPRESSIVE DISORDER NEC	30000	ANXIETY STATE NOS	05 - 311 - DEPRESSIVE DISORDER NEC	02 - 30000 - ANXIETY STATE NOS
225098371	311	DEPRESSIVE DISORDER NEC	311	DEPRESSIVE DISORDER NEC	05 - 311 - DEPRESSIVE DISORDER NEC	05 - 311 - DEPRESSIVE DISORDER NEC
237469534	30000	ANXIETY STATE NOS	30981	PROLONG POSTTRAUM STRESS	02 - 30000 - ANXIETY STATE NOS	06 - 30981 - PROLONG POSTTRAUM STRESS
229156524	311	DEPRESSIVE DISORDER NEC	30981	PROLONG POSTTRAUM STRESS	05 - 311 - DEPRESSIVE DISORDER NEC	06 - 30981 - PROLONG POSTTRAUM STRESS

Re: Create tabulation of two unique values (like times-table of old)

Tom — Mon, 13 Jan 2014 19:34:37 GMT

How would it make any difference if the DIAG variables were character?

Re: Create tabulation of two unique values (like times-table of old)

alawton — Mon, 13 Jan 2014 19:41:57 GMT

Right again. Forgive me for misunderstanding the evaluation. Very helpful.