You could also try the retain statement and/or a Cartesian join /* Thanks RW9 for the data step */ data have;   week=1; key=1; product="AA"; val=234; output;   week=1; key=2; product="BB"; val=45; output;   week=2; key=1; product="AA"; val=98; output;   week=3; key=1; product="AA"; val=23; output;   week=3; key=2; product="BB"; val=12; output;   week=4; key=2; product="BB"; val=5; output; run; /* Find the individual weeks */ proc sql; create table dates as select distinct week from have; quit; /* Find the individual keys */ proc sql; create table keys as select distinct key from have; quit; /* Find the individual products */ proc sql; create table products as select distinct product from have; quit; /* Cartesian join to generate a template */ proc sql; create table date_template as select a.week   , b.key   , c.product from dates as a   , keys as b   , products as c ; quit; proc sort data=date_template; by key product week; run; proc sort data=have; by key product week; run; data want; merge date_template   have; by key product week; /* Retain the last value assigned */ retain val1; /* Reset the value is the product is different */ if product ne lag1(product) then val1 = 0; /* If val not missing assign val to val1 */ if val ne . then val1 = val; run; ... View more