If you don't have big table, try SQL.   data have; infile datalines dlm=',' dsd truncover; input ID Returns year; datalines; 1,1.01,1968 1,1.04,1969 1,1.001,1970 1,1.005,1971 1,1.02,1972 1,1.02,1973 1,1.06,1980 1,1.06,1981 1,1.0014,1982 run; proc sql; select *,(select mean(returns) from have where id=a.id and year between a.year-2 and a.year) as rolling_mean from have as a; quit; ... View more

PROC EXPAND.    http://blogs.sas.com/content/iml/2016/01/27/moving-average-in-sas.html ... View more

Updated: data have; infile datalines dlm=',' dsd truncover; input ID Date:anydtdte. Returns Delisting_return month newmonth year; format date date9.; datalines; 1,1967-10-28,1.025,,10,07,1967 1,1967-11-28,1.026,,11,08,1967 1,1967-12-28,1.027,,12,09,1967 1,1968-01-28,1.01,,1,10,1968 1,1968-02-28,1.04,,2,11,1968 1,1968-03-28,1.001,,3,12,1968 1,1968-04-28,1.005,,4,01,1968 1,1968-05-28,1.02,,5,02,1968 1,1968-06-28,1.02,,6,03,1968 1,1968-07-28,1.06,,7,04,1968 1,1968-08-28,1.06,,8,05,1968 1,1968-09-28,1.0014,1.0014,9,06,1968 ; run; data want (drop=i fst_date fst_newmonth); set have; by id; retain fst_date fst_newmonth; if last.id and month(date) ne 12 then do; fst_date=date; fst_newmonth=newmonth; do i=0 to 12-newmonth; date=intnx('month',fst_date,i); if i > 0 then returns=.; month=month(date); year=year(date); delisting_return=.; newmonth=fst_newmonth+i; output; end; end; else output; run; ... View more

Instructions here: https://communities.sas.com/t5/SAS-Communities-Library/How-to-create-a-data-step-version-of-your-data-AKA-generate/ta-p/258712 will show how to turn an existing SAS data set into data step code that can be pasted into a forum code box using the {i} icon or attached as text to show exactly what you have and that we can test code against. ... View more

Create the dataset with the dates and values you want and insert those (via proc append) and the sort to get your desired order.  ... View more

Ordinarily you would get no answers to this question becuase you did not provide sample data for SIZE1, for FF, and for your desired outcome.   But I believe I am familiar with the particular data sources you are using.   You have data in SIZE1 sorted by permno/date in dataset size1, right?  I guess it's CRSP monthly data, where the date is the last trading day of the month.   You want to merge it with FF data, which I believe is also monthly, but is only a single series (i.e. no permno) and apparently uses last calendar date of the month.  If that's the case, then it's best to load the FF into a lookup table (called a hash object below) and just retrieve the needed values by means of hash lookup.  No need for all your sorting.     data need; set size1; if _n_=1 then do; if 0 then set FF; declare hash ffdata(dataset:'ff'); ff.definekey('date'); ff.definedata(all:'Y'); ff.definedone(); end; date=intnx('month',date,0,'end'); rc=ffdata.find(); date60=intnx('month',date,-60,'end'); run;     Notes: The hash object Is automatically "retained". That's why it is generated and populated only under the "IF _N_=1" do group.  You only need to create the hash object once.  "IF _N_=1" means to do this only during the first iteration of the data step (which usually means when processing the first observation). The "if 0 then set ff" statement tells sas to make entries in the program data vector for all the FF vars, but don't bother actually reading in the FF data. The hash object FFDATA, in this case, just takes in all the FF data:  declare hash ffdata(dataset:'FF') .  The are many attributes of the hash object that I won't go into, but are particulary useful for your task. It is "keyed" (indexed) on the date var:    ffdata.definekey('date') The data to be retrieved via that key is all the vars in FF:   ffdata.definedata(all:'Y').  If you only want some of the vars (say vars a b and c), you could just change the declare statement to   declare hash ffdata(dataset:'FF (keep=date a b c)'); Getting end-of-month dates.  You don't need date=intnx('month',date,1)-1  You can use the alignment parameter 'end", as in date=intnx('month',date,0,'end'). rc=ffdata.find() says to get the FF data for the date of the incoming SIZE1 observation.  RC=0 upon a successful lookup. Data set need now has ... Monthly data sorted by permno/date date=last calendar date of the month all the vars in SIZE1 all the vars in FF for the same month A 2nd date variable DATE60, which has the date for 60 months prior. This assumes, of course, that the FF has all the date values found in SIZE1 (after end-of-month adjustment).  Otherwise some FF values will be erroneous in the progarm above.  You would need to explicitly set them to missing whenever RC^=0.  So take a look at NEED before you run the program below to generate WANT.   Now you want 60 month lagged returns. For record J of a given permno you can just find record K of the same permno where DATE60 for J = DATE for K.  That's straightforward:   data want; merge need (in=incurrent need (in=in60months keep=date ret vwretd rename=(date=date60 ret=ret60 vwretd=vwretd60)); by permno date60; if incurrent and in60months; run;   Now data WANT is still sorted by permno/date  (and also by permno/date60). It only has complete records, i.e.for which there are both current and 60-month prior data  (that's the "if incurrent and in60months"). it has 2 new vars:   ret60 an vwretd60 represented the 60 month lagged data.   You can obviously get some lagged FF vars too by adding them to the "keep="  and "rename=" parameters ... View more