Hi Michael, Great to see you online 🙂 The default join type in Enterprise Guide is the inner join, which means the rows that are returned are where there are matches from both tables. This obviously isn't what you are after so you need to change the join type. In the Query Builder you go into the join tables window and you can double click on the join circle/Venn diagram to change this or right mouse click and select the option. I think what you are after is a full join on the tables and then you want to select the rows of interest using the filter tab. If you are joining your 3 tables at the same time, and do a full join on each of the joins then you will be getting rows that match and don't match from ALL tables. You can then use the filter tab to subset the rows based on the criteria you want. E.g. If the three values (var1,var2,var3) is in tbl_source1, are not in tbl_source2 or tbl_source3, output a subset with these observations (non-matched records). - To do this then you would need to specify in the filter tab, (tbl_Source2.var1 is missing and tbl_Source2.var2 is missing and tbl_Source2.var3 is missing) or (tbl_Source3.var1 is missing and tbl_Source3.var2 is missing and tbl_Source3.var3 is missing) or you can do the subsetting on the values of the columns from tbl_Source2 and tbl_Source3 that are missing (X and Y), if it is those values that determine non-matches. Similarly for your other queries, look at using an outer join type with filtering. Here's an EG 4.1 paper that you might find useful as a reference: http://www.scsug.org/SCSUGProceedings/2010/Schlechte/Using_Enterprise_Guide_for_Table_Joins_and_Sets.pdf As you mentioned you have done Prog 1, then I suggest to look at page 10-85 from the Prog 1 course notes where they have an example of creating multiple data sets when merging multiple tables using the in= data set options and subsetting if statement. If your data has many-to-many observations though, its probably best to use the EG query builder or write PROC SQL as the way the data step processes observation at a time will probably not produce what you want for a many-to-many merge (see pages 10-86 & 10-87 where data step is compared to sql). Hope this helps. Cheers, Michelle
... View more
