set counter

[Steelers_In_DC]

I am using a counter like this:

data test;

set data;

by power_id record_date;

cure_count+1;

if first.power_id then count= 1;

if default ne '' then count= 0;

There is a facility_id that roles up to power_id.  Is there a way to keep facility_id in this dataset but not count by it?

Here is what I have:

power_id	facility_id	record_date	cure_count
1	1	1/1/2001	1
1	2	1/1/2001	2
1	3	1/1/2001	3
1	1	2/1/2001	4
1	2	2/1/2001	5
1	3	2/1/2001	6

Here is what I want:

power_id	facility_id	record_date	cure_count
1	1	1/1/2001	1
1	2	1/1/2001	1
1	3	1/1/2001	1
1	1	2/1/2001	2
1	2	2/1/2001	2
1	3	2/1/2001	2

[LinusH]

Do cure_count+1; only when facility_id = 1?

[Steelers_In_DC]

I'm not sure I understand your solution.  This is for a table that has millions of power_id's and facility_id's.

[RW9]

Hi,

data want;

     set have;

     by power_id facility_id;

     retain cure_count;

     if first.factility_id then sum(cure_count,1);

run;

[LinusH]

From you sample it looked like facility restarts its count, each date? Your sample may not represent the real complexity.

Perhaps you could use first.record_date instead.

[Tom · Accepted solution]

Why did you bump the counter when you did?  Was it because of the change in date?

data want ;

  set have ;

  by power_id record_date ;

  if first.power_id then cure_count=0;

  if first.record_date then cure_count+1;

run;