To do the "remerge" in one step, use a double DO loop: data have; input ID $ visit_order num_stitches num_removed; datalines; X 1 4 . X 2 . 2 X 3 3 . X 4 . 3 X 5 . . A 1 5 . A 2 . . A 3 3 . B 1 2 . B 2 4 . B 3 . 3 B 4 12 . ; data want; do until (last.id); set have; by ID notsorted; retain total_stitches; if first.id then total_stitches = 0; total_stitches + num_stitches; if num_removed ne . then num_never_removed = total_stitches - num_removed; total_stitches + (-num_removed); end; patient_level = sum(0,num_never_removed); do until (last.id); set have; by ID notsorted; retain total_stitches; if first.id then total_stitches = 0; total_stitches + num_stitches; if num_removed ne . then num_never_removed = total_stitches - num_removed; else num_never_removed = .; output; total_stitches + (-num_removed); end; run; Note that I corrected the last observation of your example data. ... View more

Just use a SUM statement. data have; input ID $ visit_order num_stitches expected; DATALINES; A 1 5 5 A 2 . 5 A 3 3 8 B 1 2 2 B 2 4 6 B 3 . 6 B 4 12 18 C 1 . . C 2 . . D 1 . . D 2 2 2 D 3 . 2 ; data want; set have; by id; if first.id then total_stitches=.; total_stitches + num_stitches; run; But why are you setting total_stitches to missing instead of zero when first visit does not have any stitches?  If you would rather the total be zero in those cases just change the value that is assigned for the first observation for the patient. ... View more

@awardell wrote: Thanks for your speedy response. I have put this in SAS, and this gives me the gender by ethnicity table. I'm guessing I will need to do something similar to this to get the tables with all three variables:    proc tabulate data=gender_p; class race ethnicity gender; table race, ethnicity*gender run;   Thanks so much!!! It depends on how you want to look at things. One thing it may be helpful to know is that Proc Tabulate supports multiple table statements. Second is the , separates dimensions (page, row and column) and * nests variables and statistics within a dimension: so you might look at proc tabulate data=gender_p; class race ethnicity gender; table race, ethnicity*gender ; table race, ethnicity, gender ; table race * ethnicity, gender ; table race * ethnicity * gender ; table race ethnicity gender, race ethnicity gender ; run; for a variety. The N statistic is assumed for class variables is no other is requested. There are a number of percentages that could be requested as well. ... View more