Subtract first from second rows and put it in a new column

[million]

Hi,

I have a data that has three variables namely Department, Totalsales and Date. I have many departments. Each day, new sales adds to the the Totalsales column and shows the total until that specific day. I want to know how many new sells there are each day by subtracting one day's totalsales from the next day's totalsales and put it in a new column. Therefore at last I will have four columns (Department, Totalsales, Date, and newsells). I sorted the data by Department and Date. I tried to work using do loop but lost on the way. Can anybody assist me with this. I am posting a hypothetical data similar to my data here.

Thank you

Data All;
Input Department$ Totalsales Date DDMMYY8.;
format Date yymmdd10.;
cards;
D 45 01/03/19
D 47 02/03/19
D 51 03/03/19
D 55 04/03/19
B 35 01/03/19
B 37 02/03/19
B 43 03/03/19
B 42 04/03/19
E 59 03/03/19
E 64 04/03/19
M 45 01/03/19
M 50 02/03/19
M 55 03/03/19
M 56 04/03/19
M 59 05/03/19
M 59 07/03/19
M 59 06/03/19;
run;

I want new sales for each department in the first column.

[mkeintz · Accepted solution]

A loop would be useful if you had to do the same process to a set of variables.  But you have only one variable (totalsales), and you want to compare successive observations within the data set:

Data All;
  Input Department$ Totalsales Date DDMMYY8.;
  format Date yymmdd10.;
cards;
D 45 01/03/19
D 47 02/03/19
D 51 03/03/19
D 55 04/03/19
B 35 01/03/19
B 37 02/03/19
B 43 03/03/19
B 42 04/03/19
E 59 03/03/19
E 64 04/03/19
M 45 01/03/19
M 50 02/03/19
M 55 03/03/19
M 56 04/03/19
M 59 05/03/19
M 59 07/03/19
M 59 06/03/19
run;
data want;
  set all;
  by department notsorted;
  newsales=dif(totalsales);
  if first.department then newsales=.;
run;

The DIF function is defined as   dif(x)=x-lag(x).  The "if first.department …" statement tests for the start of a new department, and set newsales to missing, thereby avoiding contamination of the start of one department with a lagged value from the prior department.

The "notsorted" option is because you seem to have your data observations grouped by department, but not in ascending order.

[million]

Thank you so much. This works!

Just a followup question. What if we want the the start of a new department column to be the same as the totalsales value in that column? Not a missing value.

[mkeintz]

Sorry for the delayed response - been offline. I leave the answer to your question as a "homework" assignment.