Cheema,   You have received a solution (SQL)  for almost same problem with a different subject line from KSHARP. You want an efficient solution now. Tell us your experiments with SQL solution and  any other method you tried. Give some results of run-times and memory issues.   regards, DataSp ... View more

You may remember: Since your dependent variable, BMI, is computed using Height and Weight, the inclusion of the latter as independent variables, will be highly correlated. ... View more

You have several ways to do your job. But you indicate that you are not getting the expected output.    In your sample data set, you have not provided one condition to be checked. You want the TEMP with a missing value for the first and last date of the ZIP be retained as such. So I have modified your input data for checking the condition :   data have; INFILE CARDS EXPANDTABS; input date date9. zip temp; cards; 01Jan2005 90001 . 02Jan2005 90001 51 03Jan2005 90001 52 04Jan2005 90001 49 05Jan2005 90001 48 06Jan2005 90001 . 07Jan2005 90001 50 08Jan2005 90001 51 09Jan2005 90001 52 10Jan2005 90001 49 11Jan2005 90001 54 12Jan2005 90001 60 13Jan2005 90001 50 14Jan2005 90001 . 15Jan2005 90001 52 16Jan2005 90001 49 17Jan2005 90001 . 18Jan2005 90001 60 19Jan2005 90001 50 20Jan2005 90001 . 01Jan2005 90002 60 02Jan2005 90002 51 03Jan2005 90002 52 04Jan2005 90002 52 05Jan2005 90002 52 06Jan2005 90002 . 07Jan2005 90002 . 08Jan2005 90002 . 09Jan2005 90002 . 10Jan2005 90002 . 11Jan2005 90002 . 12Jan2005 90002 . 13Jan2005 90002 . 14Jan2005 90002 . 15Jan2005 90002 52 16Jan2005 90002 49 17Jan2005 90002 52 18Jan2005 90002 49 19Jan2005 90002 48 20Jan2005 90002 . ; run; Next, we scan the DATA SET to get the SMALLEST and the BIGGEST Date to SIZE the Array to be used in the following Program. If you know these before hand, you can skip this step but you need to create Macro variables for them (bDate, eDate). The array is sized for the biggest group. So, small groups will also fit into it.   data _null_; retain eDate 0 sDate 999999999; set have end = eof; if date > eDate then eDate = date; if date < sDate then sDate = date; if eof then do; call symputx('sDate', sDate); call symputx('eDate', eDate); end; run; The program uses two DO-LOOPs. This do-loop is known as DoW-loop. In the first do-loop, TEMP for each DATE is saved in a _temporary_ array. Only one Zip at a time is processed. We save the first DATE and the last DATE to check your boundary conditions. In the second do-loop, the same group is scanned again and checked for your requirements. Here is the program:   options fullstimer; data want; array k[&sDate:&eDate] _temporary_; do until(last.zip); set have; by zip; k[date] = temp; if first.zip then startD = date; if last.zip then endD = date; end; do until(last.zip); set have; by zip; if (date = startD & temp = .) | (date = endD & temp = .) then temp_new = .; else if not missing(temp) then temp_new = temp; else do; tf = .; td = .; do m = date+1 to date+7; * going forward; if k[m] > 0 then do; tf = k[m]; leave; end; end; do m = date - 1 to date - 7 by -1; * going backward; if k[m] > 0 then do; td = k[m]; leave; end; end; if missing(tf) or missing(td) then temp_new = .; else temp_new = (tf + td) * 0.5; end; output; end; keep date zip temp temp_new; run; Here is the OUTPUT generated by this program:   Obs date zip temp temp_new 1 16437 90001 50 . 2 16438 90001 51 51.0 3 16439 90001 52 52.0 4 16440 90001 49 49.0 5 16441 90001 48 48.0 6 16442 90001 . 49.0 7 16443 90001 50 50.0 8 16444 90001 51 51.0 9 16445 90001 52 52.0 10 16446 90001 49 49.0 11 16447 90001 54 54.0 12 16448 90001 60 60.0 13 16449 90001 50 50.0 14 16450 90001 . 51.0 15 16451 90001 52 52.0 16 16452 90001 49 49.0 17 16453 90001 . 54.5 18 16454 90001 60 60.0 19 16455 90001 50 50.0 20 16456 90001 51 . 21 16437 90002 60 . 22 16438 90002 51 51.0 23 16439 90002 52 52.0 24 16440 90002 52 52.0 25 16441 90002 52 52.0 26 16442 90002 . . 27 16443 90002 . . 28 16444 90002 . 52.0 29 16445 90002 . 52.0 30 16446 90002 . 52.0 31 16447 90002 . 52.0 32 16448 90002 . 52.0 33 16449 90002 . . 34 16450 90002 . . 35 16451 90002 52 52.0 36 16452 90002 49 49.0 37 16453 90002 52 52.0 38 16454 90002 49 49.0 39 16455 90002 48 48.0 40 16456 90002 60 . You can compare all the methods and see how FULLSTIMER gives Time and Memory. Enjoy. ... View more

I have not used Proc SurveySelect. This is my small advice.   For every case there can be 100s of Controls. Before decide select a Control, decide on the characteristics of patients which have direct impact on outcome variable( Complication X). Usually, Age, Gender, any residential characteristics as an indirect indicator of their economic status, plus any other potential variables that may be associated with X. Exact match on these with Case and Control may not be feasible. So some ranges in variables have to be accepted.  You match on Gender, but Age will not. For example, Age of Case plus or minus 1 may be considered for selecting the Control. This relaxing conditions be fixed at this stage. The important note is that selecting more than 3 Controls per Case will not be profitable for Case-Control studies, less is better. Let the number of Controls be 2.   Select a case with Complication X.      Note the Age, Gender, Other Variables you have decided. Find those without Complication X.     Match with Age, Gender, Other Vars. Count the number of them.     If 1 control is found, choose the patient.    If you have more than say, 10, then choose 2 of them at random. You may use FLOYD Algorithm for this.    If no Control is found for a selected Case, relax the matching criteria as decided in advance before the selection process.      ... View more

Can you provide sample data for two zip codes each having 8 - 10 observations to check for 7-day window? Show the resulting output you need from out of that data. ... View more

Not sure without seeing the desired data set ( WANT). Simple  use of "????" does not show what you need. Will it work if you use a DO - satement like:   if first.id then DO;       date_r1=date;       date_r2 = date; /* NOT SURE WHAT IS TO BE ASSIGNED */       date_r3 = date; /* NOT SURE WHAT IS TO BE ASSIGNED */ END; ... View more

(10202008AB_RANK00325)   You want a numeric from the above. I only know that a NUMERIC contains only digits from 0 to 9.  Can you show how  you want that numeric should look like?  ... View more

BerndSE: You asked:    do i need the "if 0 then set sample" from your code? Consider:   if 0 then set sample; declare hash h(dataset:"sample", ordered: 'Y');   When you are loading the Data set ( SAMPLE ) in Bulk into the HASH TABLE, you need   if 0 then set sample;   to IMPACT the Program Data Vector (PDV). If that statement is not there, SAS does not consult the metadata of the data set by itself and will not succeed in defining the hash table.   For curiosity, drop the statement, run the program. This is how you learn the hash objects !   ... View more