Proc sql: new and continue customers based on look

问题:

I have following data:

wei 01feb2018    car
wei 02feb2018    car
wei 02mar2019    bike
carlin 01feb2018 car
carlin 05feb2018 bike
carlin 07mar2018 bike
carlin 01mar2019 car

I want to identify new and continue customers, if a customer had no purchase in last 12 months then it will become a new customer. Required output be like

wei 01feb2018    car     new
wei 02feb2018    car     cont.
wei 02mar2019    bike    new
carlin 01feb2018 car     new
carlin 05feb2018 bike    cont.
carlin 07mar2018 bike    cont.
carlin 01mar2019 car     new

Now if a customer has purchased any item in the same month for ex -customer a purchased car on 01jan and bike on 15jan then I want two classify customer a as new for Jan for one report and in another report I want customer a as both new and continue.

I'm trying but not getting the logic -

proc sql;
select a.*,(select count(name) from t where intnx("month",-12,a.date) >= 356) 
as tot
from t a;
Quit;

回答1:

You can use retain:

proc sort data=test out=test2;
  by name  type date;
run;

data test2 ;
set test2;
 retain retain 'new';
 by name  type date;
   if first.type then retain='new';
   else retain='con';
run;

proc sort data=test2 out=test2;
  by name  date; 
run;

Output:

+--------+-----------+------+--------+
|  name  |   date    | type | retain |
+--------+-----------+------+--------+
| carlin | 01FEB2018 | car  | new    |
| carlin | 05FEB2018 | bike | new    |
| carlin | 01MAR2019 | car  | con    |
| wei    | 01FEB2018 | car  | new    |
| wei    | 02FEB2018 | car  | con    |
| wei    | 02MAR2019 | bike | new    |
+--------+-----------+------+--------+

回答2:

You appear to want two different 'status' variables, one for continuity over prior year and one for continuity within month.

In SQL an existential reflexive correlated sub-query result can be a case test for rows meeting the over and within criteria. Date arithmetic is used to compute days apart and INTCK is used to compute months apart:

data have; input
customer $ date& date9. item& $; format date date9.; datalines;
wei     01feb2018  car
wei     02feb2018  car
wei     02mar2019  bike
carlin  01feb2018  car
carlin  05feb2018  bike
carlin  07mar2018  bike
carlin  01mar2019  car
run;

proc sql;
  create table want as
  select *,
    case
      when exists 
      (
        select * from have as inner 
        where inner.customer=outer.customer
          and (outer.date - inner.date) between 1 and 365
      ) 
      then 'cont.'
      else 'new'
    end as status_year,
    case
      when exists 
      (
        select * from have as inner 
        where inner.customer=outer.customer
          and outer.date > inner.date
          and intck ('month', outer.date, inner.date) = 0
      ) 
      then 'cont.'
      else 'new'
    end as status_month
  from have as outer
  ;
quit;