An example of @PGStats comment about the syntax of the IN list you are using in the version 2:
data example;
do i= 1 to 10;
inlist= ( i in (2-5,7, 8-10));
rightlist = ( i in (2:5,7, 8:10));
put i= inlist= rightlist=;
end;
run;
Inlist and Rightlist will have a value of 1 when I is found in the list. Note that only the first of values in the hyphenated are matched for Inlist.
With a smaller set of variables but the same type of code as your version 1 a cannot duplicate getting "all values of 3" set to .P.
So either the code you ran is different than shown or perhaps you have other code assigning values of P, or copying from one of the affected variables.
Since you did not show 1) an entire data step, 2) the log entry from running the code and 3) provided no test data it is hard to determine what may have happened to set "all values of 3".
array temp{5} 3 3 6 7 3; /* so temp is a temporay array that disappears along with its contents after run execution */
A temporary array of that sort would look like
array t{5} _temporary_ (3,3,6,7,3);
Caution: temporary arrays are not reset at each iteration of the data step. So if change values inside the array they will be the new values for the next record(s).
From the values I am seeing I would hazard guess that something like certain question responses are being set where the same "not really useful for analysis value" like "Don't know" "Refused" or similar is set to .P.
I tend to that with custom informats when reading data most of the time. So none of this scattered array indexing is needed.
Or create three separate arrays for the purpose of recoding.
Variables can be assigned to multiple arrays if needed. So if your vrra is doing something else other than recoding then recode in separate arrays for groups of variables prior to the other stuff.
