how to calulcate the difference between two rows

[manya92]

Hi all. My dataset looks like the one below. I need to make a new variable called modification based on the variable amount. if the difference between the subsequent amounts for the same patientID is at least 15 then the dose is modified else not. How to achieve this in SAS?

Obs PatientID sub_dates amount1234567891011121314151617181920

F012DC0F325EF	12/31/2014	250
F012DC0F325EF	01/14/2015	337
F012DC0F325EF	01/28/2015	337
F012DC0F325EF	02/16/2015	337
F012DC0F325EF	03/02/2015	337
F012DC0F325EF	03/26/2015	337
F012DC0F325EF	04/13/2015	337
F012DC0F325EF	04/27/2015	337
F012DC0F325EF	05/13/2015	337
F012DC0F325EF	05/28/2015	337
F012DC0F325EF	06/10/2015	337
F03254A1ADC26	01/11/2017	350
F03254A1ADC26	01/30/2017	350
F03254A1ADC26	02/13/2017	350
F03254A1ADC26	03/01/2017	350
F03254A1ADC26	03/22/2017	350
F03254A1ADC26	04/04/2017	350
F03254A1ADC26	04/18/2017	350
F0577A8783BD9	06/25/2014	330
F0577A8783BD9	07/09/2014	330

[PeterClemmensen]

A few questions.. Do you have SAS/ETS? Also, is 'difference' both positive and negative? Should the dose be modified if the subsequent value is both 16 higher/lower than the current observation?

[ballardw]

As a hint for asking questions in the future:

1) provide just enough data to illustrate the issue/ logic/ question. The last 6 or 7 observations for patient F012DC0F325EF are not actually adding any information to the use case

2) Show what your expected output given the input example would be. For instance you do want a new variable with the difference of the amount? Or a flag that indicates "modify" when out of range.

3) As @PeterClemmensen asked, be sure to clearly indicate clearly which values meet your objective. It would be best if your input data showed both declines and increases past the threshold so that the example output would clarify whether not a specific observation meets the rule.

4) Ideally your example data should be in the form of data step code that allows us to test code against values. And you should provide the data step because if we make one with similar data we might make a choice that does not work for your data. For instance I would make sure that your date value is a SAS date value. But we have lots of questions about dates on this forum where it turns out the actual variable for a date is character. So solutions using actual date values "do no work" for a posters question because they did not describe their data accurately. A data step with the same variable names, types, lengths and formats means that we are on the same page for the data.