How to lag by region, month, & year?

[jlajla]

Hello,

I am attempting to lag a variable by region, month, and year.  Essentially, I need to create a new variable that is the lagged mean value (by year and month) for a neighboring region.

I have this:

region month year mean

1 1 00 5

1 2 00 6

1 3 00 4

1 1 01 4

1 2 01 5

1 3 01 5

2 1 00 4

2 2 00 5

2 3 00 3

2 1 01 3

2 2 01 5

2 3 01 2

and I need:

region month year mean neighbor_mean_lag

1 1 00 5 .

1 2 00 6 .

1 3 00 4 .

1 1 01 4 10

1 2 01 5 13

1 3 01 5 16

2 1 00 10 .

2 2 00 13 .

2 3 00 16 .

2 1 01 3 .

2 2 01 5 .

2 3 01 2 .

(Note neighbor_mean_lag equals the neighboring region's monthly values for the previous year.)

I am able to create a lagged variable for the same region using the below code. However, I haven't been successful at creating the lagged variable for a neighboring region.

DATA lag3;
     SET lag2;
      BY region month year;
      lag1_mean=lag(mean);
      if first.month and first.year then do;
         lag1_mean=.;
            end;
RUN;

Thank you for your help!

[PGStats]

A self join will be simpler than using LAG

data have;
input region month year mean;
datalines;
1 1 00 5
1 2 00 6
1 3 00 4
1 1 01 4
1 2 01 5
1 3 01 5
2 1 00 4
2 2 00 5
2 3 00 3
2 1 01 3
2 2 01 5
2 3 01 2
;
data have1;
set have; by region notsorted;
reg + first.region;
run;

proc sql;
create table want as
select 
    a.region,
    a.month,
    a.year,
    a.mean, 
    b.mean as neighbor_mean_lag
from 
    have1 as a left join
    have1 as b 
    on  b.reg = a.reg+1 and 
        b.year = a.year-1 and 
        b.month = a.month
order by a.reg, year, month;
quit;

[Ksharp]

looks like it is not a mean, it is a  MEAN variable value for the next region. I would try Hash Table.