Perhaps this is less complicated. Read all the records for a given id once to build an array (VIS) of visit values. Each element of the array contains the visit number for the corresponding group of records. Each group is a sequence of identical visit values. For ID 101, there are seven groups with values {1, ., 2, ., 2, 3, .}.
The reread (and output) the same records. If a record has a value of . and is in group g, then check whether groups g-1 and g+1 have identical values. If they do, then fillin in that value.
So if group g has a visit value of . and groups g-1 and g+1 have matching values, then replace.
data want (drop=_:) ;
array vis{0:20} _temporary_;
do _g=1 by 1 until (last.id);
do until (last.visit);
set one;
by id visit notsorted;
end;
vis{_g}=visit;
end;
_ng=_g;
do _g=1 to _ng;
if vis{_g}=. and vis{_g-1}=vis{_g+1} then _fillvis=vis{_g-1};
else _fillvis=.;
do until (last.visit);
set one;
by id visit notsorted;
if visit=. then visit=_fillvis;
output;
end;
end;
run;
... View more
