Rather than use ErrorHandler here, consider writing your logic into a function that acts on a scalar struct. This function would check the input struct's Num field size to determine whether to return a mean calculation or 0. Here is an example:
function v = meanMyStruct(s)
sNum = s.Num;
if size(sNum,1) >= 20
v = mean(sNum(1:20,3));
else
v = 0;
end
Now you can use this function with ARRAYFUN on your struct array:
s = struct('Num',{rand(20,5), rand(19,5), rand(21,5)});
out = arrayfun(@meanMyStruct, s)
