I appreciate your feedback. Creating the flag is an interesting solution. How would I adjust the code so it will flag day 2 as a high?   would these minor changes in step1 and step 2 work? sign=sign(coalesce(dif(temp),1);   if last.sign and run>1 then do;   I don't think those changes will work if the next sequence of the data is below   I am not trying to be contradictory. I do want to flag the end of a run of increasing (decreasing) days, but I am allowing for a whipsaw of temperatures. For example, if the next sequence of temperatures was: Day       Temp 16             78 17             80 18             79 19             81 20             80 21             82 22             81   the output would flag: day17- High, day18- Low, day19- High, day20- Low, day21- High ... View more

Thank you for your input. However, I would like to use the following definitions to distinguish a local high (low): Local high: if temp(yesterday) > temp(today) and temp(yesterday) > temp(2 days ago)  Local low: if temp(yesterday) < temp(today) and temp(yesterday) < temp(2 days ago)   your method has any observation of high > 2 as a local high. This would mean there was a local high on days 8 and 9, however, day 8 cannot be a local high because temp(day8) > temp(day7) but temp(day8) is not great than temp(day9).   Would it be easiest to create a loop in SAS that evaluated 3 temperatures of consecutive days? if temp(yesterday) > temp(today) and temp(yesterday) > temp(2 days ago) keep, otherwise delete; or  if temp(yesterday) < temp(today) and temp(yesterday) < temp(2 days ago) keep, otherwise delete; I won't have the high_i (low_i) labels, but I will have the dates and temperature values of all the local highs (lows) according to definition I created ... View more

Thank you for your input. I agree generating a new variable of high_i (low_i) as SAS compares the temperatures of today, yesterday, and 2 days ago is confusing.   I think the easiest way to achieve the local high (low) results, is to use a loop comparing temperatures of today, yesterday, and 2 days.  If temp(yesterday) > temp(today) and temp(yesterday) > temp(2 days ago) or If temp(yesterday) < temp(today) and temp(yesterday) < temp(2 days ago) then SAS should keep the data, otherwise SAS should delete the data.   Would I need to use a loop in SAS to use this method?   I won't have the labels of high_i (low_i), but I will be able to compare adjacent local high and low values ... View more

Sorry I will try to be more clear. I would like to compare temperature values after each day.    high_1 is found by comparing 3 consecutive days (day 3,2,1). Because temp(day 2) > temp(day1) and temp(day2) > temp(day 3) this would be the first local high. A local high cannot be found at day 1 because there is not 3 days of temperatures to compare yet (I can compare temperatures after day 3). I want the high_1 data/value to have day and temperature information.   low_1 is found by comparing 3 consecutive days (day 6,5,4). Because temp(day5) < temp(day4) and temp(day5) < temp(day6), day 5 would be considered a local low.   I thought about having SAS graph the temperatures over time, then compare the change in temperature between 2 days (the slope of the change). When the slope transitions from positive to negative, that would be a local high (and local low when the slope transitions from negative to positive). Could SAS label each high (low) value (using the change in slope coefficient)?   Another solution I was thinking about was having SAS drop values when the following is not true: If temp(yesterday) > temp(today) and temp(yesterday) > temp(2 days ago) or If temp(yesterday) < temp(today) and temp(yesterday) < temp(2 days ago) This would create a dataset with only local high and low values [along with information about which day was a high(low)]   However, I was hoping there was a code to generate high_i (low_i) if my dataset had a decade worth of temperature data, rather than deleting values from the original data set to only include local high(low) values   Thanks ... View more

Here is an example of data: Temperature is the average of the high and low for the day Day        Temperature(*F) 1                     74 2                     76 3                     75 4                     73 5                     72 6                     74 7                     77 8                     78 9                     80 10                   79 11                   78 12                   76 13                   75 14                   76 15                   77   The output I want is: high_1: on day 2 @76* low_1: on day 5 @72* high_2: on day 9  @80* low_2: on day 13 @75*    If my dataset was over 1000 days of data, I want to calculate all the local high_i and low_i using a 3 consecutive day comparison of temperatures ... View more