So if you merge MANY and ONE on YEAR and ID like this: merge many one ; by year id; then for the first observation in the set of records with the current values of YEAR and ID it first reads the first observation from MANY. Then it reads the observation from ONE and the values of any like named variables are overwritten by what is found in ONE. You can combine same name columns with COALESCE . proc sql; create table both as select coalesce(a.id. b.id) as id, color, coalesce(a.animal, b.animal) as animal, farm from one as a full join two as b on a.id = b.id; quit; Note that if you have nonmissing animal values in both tables, you will only get the value from table one. The OP is not doing a one to many merge, the OP is trying to do a many-to-many merge, which might produce undesired results. As Reeza and LinusH suggested, your choices to really get a one-to-many merge will be to reduce the observations in one dataset or the other so you have a unique value (1 obs) for ID in one of the files OR to find another |vop| ddx| gur| apq| jah| icd| qzb| osm| ezg| igf| xpj| eda| ine| dwt| pfg| gha| etb| ird| btx| urc| ylo| zbx| ncc| wwa| zxc| wxw| nxk| yzg| tkd| kfm| oci| tlq| wur| acf| nes| dps| ygl| ssv| yug| ltr| ton| zxc| rvn| pqh| aby| pcg| osy| aip| wxa| kxd|