On Fri, Jun 10, 2011 at 3:26 PM, Bolotin Yevgeniy
> The SUM function (and probably other aggregate functions) in PROC SQL
> change from row-level to column-level based on context
> Is it possible to force it to be one vs the other?
> (as a thought experiment: I want to get a row-level sum of non-missing
> values, but the variable list is a dynamically generated macro parameter
> and may only have one variable in it)
> Sample data demonstrating this is below.
> for a fun time, call sum(sum(a,b,c))!
> data _z;
> input a b c;
> 1 2 3
> 4 . 6
> proc sql;
> sum(a,b,c) as sum_row,
> sum(a) as sum_col
> from _z
> sum_row sum_col
> 6 5
> 10 5