Hi there,   This should be pretty close to what you are looking for. Note that for "Q0002" and "Q0006", multiple summary statistics are returned, so you will need to use another summary function such as MIN(), MAX() or AVG(). In this example, I used AVG, but you can use MIN or MAX if that better suits your needs.    PROC SQL; /* CREATE TABLE backTestGroup2 AS */ CREATE TABLE backTestGroup3 AS SELECT Variable length=20 , COUNT(DISTINCT CATX('|', Mean, Stddev, Center, Slope, Mean)) AS num_distinct_values , AVG(Mean) AS Mean_Mean , AVG(Stddev) AS Mean_Stddev , AVG(Center) AS Mean_Center , AVG(Slope) AS Mean_Slope , CASE WHEN calculated num_distinct_values > 1 THEN '<== Note: summary statistic is an average (more than one unique row was returned)' ELSE '' END AS remark length=100 FROM (SELECT DISTINCT TRANSTRN(Variable, "_", "") AS Variable, Mean, Stddev, Center, Slope FROM ozMeans UNION SELECT DISTINCT TRANSTRN(Variable, "_", "") AS Variable, Mean, Stddev, Center, Slope FROM backTestGroup2 WHERE TRANSTRN(Variable, "_", "") NOT IN (SELECT Variable FROM ozMeans)) As InnerQry GROUP BY Variable ORDER BY 1; QUIT;   Here is the output:  (edited: updated image; old output was incorrect)     Enjoy.    -- hbi ... View more

This could be done in fewer steps, but I used multiple PROC SQL statements to show what is done at each step to arrive at the final results.   [Edited: you may want to change "EXCEPT ALL" to "EXCEPT". If your data contains multiple rows with identical values (e.g. three observations in Table_A that all have "1001 / ABCD / Good"), "EXCEPT" should give you better results.]   DATA Table_A; LENGTH Code 8 Test $10 Value $20; INPUT Code Test $ Value $; DATALINES; 1001 ABCD Good 1001 ABCD Bad 1001 ABCD Above_average 1001 ABCD Average ; RUN; DATA Table_B; LENGTH Code 8 Test $10 Value1 $20; INPUT Code Test $ Value1 $; DATALINES; 1001 ABCD Bad 1001 ABCD Good 1001 ABCD . 1001 ABCD Below_average ; RUN; PROC SQL; CREATE TABLE Table_A_Not_B AS SELECT Qry.* , monotonic() AS seq_num FROM (SELECT * FROM Table_A(WHERE=(Value IS NOT NULL)) /* Change "EXCEPT ALL" to "EXCEPT" */ EXCEPT SELECT * FROM Table_B(WHERE=(Value1 IS NOT NULL))) AS Qry; QUIT; PROC SQL; CREATE TABLE Table_B_Not_A AS SELECT Qry.* , monotonic() AS seq_num FROM (SELECT * FROM Table_B(WHERE=(Value1 IS NOT NULL)) /* Change "EXCEPT ALL" to "EXCEPT" */ EXCEPT SELECT * FROM Table_A(WHERE=(Value IS NOT NULL))) AS Qry; QUIT; /* this would give you the results exactly the way you want (side-by-side) */ PROC SQL; CREATE TABLE Want_Side_by_Side AS SELECT COALESCE(T1.Code, T2.Code) AS Code , COALESCE(T1.Test, T2.Test) AS Test , T1.Value , T2.Value1 , COALESCE(T1.seq_num, T2.seq_num) AS seq_num FROM Table_A_Not_B AS T1 FULL OUTER JOIN Table_B_Not_A AS T2 ON (T1.Code = T2.Code AND T1.Test = T2.Test AND T1.seq_num = T2.seq_num); QUIT; /* alternate version: this would give you the results vertically (stacked); this approach would be more traditional in SQL because it avoids assigning and match/mismatch preference (Ballardw had pointed this out above) */ PROC SQL; CREATE TABLE Want_Stacked(drop=seq_num) AS SELECT T1.*, 'Table_A_Not_B' AS Row_Source FROM Table_A_Not_B AS T1 UNION SELECT T2.*, 'Table_B_Not_A' AS Row_Source FROM Table_B_Not_A AS T2; QUIT;   Hope this helps.   -- hbi   ... View more