this might work:
// make a column to hold results, it will be last column
dt << newcolumn("answer");
for(irow=1,irow<=nrows(dt),irow+=1,
// assumes columns 1 and 2 and last are NOT data...
// grab one row at a time...
vec = dt[irow, 3::(ncols(dt)-1)];
// get indexes (loc) of non-missing values...pick the right-most (max)
nonmissindex = max(loc(vec));
// handle possible edge case (no data)
value = if(ismissing(nonmissindex),.,vec[nonmissindex]);
// fill in the answers
dt:answer[irow] = value;
);
If you turn this into a column formula, you'll want to list the source columns explicitly for table updates and not needing dt.
dt << New Column( "answer",
formula(
vec = run_1 || run_2 || run_3 || run_4;
nonmissindex = Max( Loc( vec ) );
If( Is Missing( nonmissindex ),
. // missing if all missing
, //
vec[nonmissindex]
);
)
);
As a formula, the answer column will update when the data changes.
Craige