Calculate daily usage from cumulative usage

[smcconn5]

Hello,

I have a large data set with customer IDs and daily cumulative usage readings. I'm looking to calculate the daily usage per customer from the cumulative usage and add it as a variable in the data set. Any suggestions on how to even begin doing this? Thank you!

HAVE
ID     date     cumulative_reading     
1     10/1/19     100
1     10/2/19     150
1     10/3/19     200
2     10/1/19     100
2     10/2/19     500
2     10/3/19     900



WANT
ID     Date     Cumulative_reading     Daily Usage
1     10/1/19      100                          50
1     10/2/19      150                          50
1     10/3/19      200                              
2     10/1/19      100                          400
2     10/2/19      500                          400
2     10/3/19      900

[PaigeMiller]

data want;
    set have;
    by id;
    prev_cumulative=lag(cumulative_reading);
    if first.id then daily=cumulative_reading;
    else daily=cumulative_reading-prev_cumulative;
    drop prev_cumulative;
run;

This assumes the data is properly sorted.

Also, you might want to re-think your example, there is no logical way to get a daily of 50 for the first row. 

[Tom · Accepted solution]

Much easier if you put the value on the current record instead of the previous one.

data want;
  set have ;
  by id ;
  daily_usage=dif(cumulative_reading);
  if first.id then daily_usage=.;
run;

Results:

                           cumulative_    daily_
Obs    id          date      reading       usage

 1     1     2019-10-01        100            .
 2     1     2019-10-02        150           50
 3     1     2019-10-03        200           50
 4     2     2019-10-01        100            .
 5     2     2019-10-02        500          400
 6     2     2019-10-03        900          400

You can even decide if you want the first record's value to be missing or set it to the first value.

  if first.id then daily_usage=cumulative_reading;

Results:

                           cumulative_    daily_
Obs    id          date      reading       usage

 1     1     2019-10-01        100          100
 2     1     2019-10-02        150           50
 3     1     2019-10-03        200           50
 4     2     2019-10-01        100          100
 5     2     2019-10-02        500          400
 6     2     2019-10-03        900          400

[Kurt_Bremser]

You need to do a "look-ahead":

data have;
input ID date :mmddyy10. cumulative_reading;
format date yymmddd10.;
datalines; 
1     10/1/19     100
1     10/2/19     150
1     10/3/19     200
2     10/1/19     100
2     10/2/19     500
2     10/3/19     900
;

options mergenoby=nowarn;

data want;
merge
  have
  have (
    firstobs=2
    keep=id cumulative_reading
    rename=(id=nextid cumulative_reading=nextval)
  )
;
if id = nextid then daily_usage = nextval - cumulative_reading;
drop nextid nextval;
run;

proc print data=want noobs;
run;

Result:

                    cumulative_    daily_
ID          date      reading       usage

 1    2019-10-01        100           50 
 1    2019-10-02        150           50 
 1    2019-10-03        200            . 
 2    2019-10-01        100          400 
 2    2019-10-02        500          400 
 2    2019-10-03        900            .