Each week, there are about 5-10 questions with this respect. The simple answer is, "not this way". The only way is an extensive use of eval and that's precisely what I wouldn't do. It's like carefully avoiding the benefits of Matlab.
If it isn't too much effort, start creating your variables again and put them into an array, thus, instead of something like
myvarname=genvarname(repmat({'A'},1,5));
eval([myvarname{1} '=rand(13);'])
eval([myvarname{2} '=rand(13);'])
eval([myvarname{3} '=rand(13);'])
you better do something like
for cnt=1:900
A(:,:,cnt)=rand(13);
end
You can also preallocate and improve performance this way. Then, summation of the columns is just
The column sums are already combined row wise (if that's what you mean).
If this totally doesn't work for you because the creation of your variables took too much time, you can do something like
for cnt=1:900
columnsum(cnt,:)=eval(['sum(', myvarname{cnt} ')']);
end
But you'll have not much fun with debugging etc, so I really wouldn't recommend that.
