This should work for you. It calculates the mean gains and losses using a simple average for the first 14 daily changes, and Wilder's method for the remaining means. Hopefully, the comments in the code clearly explains the inner workings, but feel free to ask.  data rsi; set have; by stock; * compute daily change in price; prev_close = lag(close); change = close - prev_close; * determine gain or loss; if change > 0 then gain = change; else if change < 0 then loss = abs(change); * reset values when stock changes; if first.stock then do; days = 0; prev_close = .; change = .; gain = .; loss = .; sum_gain = .; sum_loss = .; mean_gain = .; mean_loss = .; end; days + 1; * save the calculated averages across iterations of data step; retain mean_gain mean_loss; * aggregate the daily gains and losses for the first 15 days (but only has 14 days of daily changes); if days <= 15 then do; sum_gain + gain; sum_loss + loss; if days = 15 then do; * simple average of first 14 values for 1st average; mean_gain = sum_gain / 14; mean_loss = sum_loss / 14; end; end; else do; * Wilder's smoothing method for remaining averages > day 15; * the mean_ values on the right side of equation are prior means; mean_gain = (13 * mean_gain + coalesce(gain, 0)) / 14; mean_loss = (13 * mean_loss + coalesce(loss, 0)) / 14; end; * calculate RSI beginning at day 15 because * 14 days of daily changes exist at this point; if days >= 15 then do; rs = mean_gain / mean_loss; rsi = 100 - 100 / (1 + rs); end; run;  I used the stock prices in sashelp.stocks to create the test data. Here is the code for that. proc sort data=sashelp.stocks out=stocks; where stock in ('IBM','Intel'); by stock date; run; * keep only 30 days of closing prices for each stock; data have; set stocks; by stock; retain days; if first.stock then days = 0; days + 1; if days <= 30 then output; keep stock date close; run; ... View more