I really don't know where to begin but I've been trying and failing for days.
The end is I need to count how many times type a happened 90 days before type b, but there cannot be any other type b's 90 days before. I can do this if I'm only looking at one instance of type b. I did then by adding a new var of the date of type b of a person's to all their other claims. I have a very crude idea how to do this for all but I'm hoping for a more elegant solution. Here are the funny parts of my data: type b can happen multiple times in the same claim so day and in that case I would treat it as one type b event. A person can also have many type b's, I think it can be as high as 150 of them. So I need to mark the b events that have no other b events 90 days before so then I can count the type a events that happened in that window. In my real data type a can be many types and I will need to have how many time a type a happened and what category the type a was.
| ID |
Claim_ID |
Claim_line |
Date |
Type |
| 1 |
1 |
1 |
11-Feb |
a |
| 1 |
2 |
1 |
11-Mar |
b |
| 1 |
2 |
2 |
11-Mar |
b |
| 1 |
2 |
3 |
11-Mar |
a |
| 1 |
3 |
1 |
11-Jun |
b |
| 1 |
3 |
2 |
11-Jun |
a |
| 1 |
3 |
3 |
11-Jun |
a |
| 1 |
4 |
1 |
3-Jul |
a |
| 1 |
5 |
1 |
6-Jun |
a |
| 1 |
6 |
1 |
9-Sep |
b |
| 2 |
1 |
1 |
8-Aug |
a |
| 2 |
2 |
1 |
10-Sep |
b |
| 2 |
3 |
1 |
30-Nov |
a |
| 2 |
4 |
1 |
11-Dec |
b |
|
