Datasets can be a bit fiddly at times.
I don't have Matlab to hand so will have to go by memory.
In my opinion, the key to working with datasets is to understand that indexing a dataset via normal indexing methods returns a dataset but indexing it via a "dot" extension can return a single column as a nondataset item, e.g.:
% Define a dataset as follows [hopefully I've got this right]
ds = dataset( {rand(10, 1), repmat({'strs'}, 10, 1) }, 'VarNames', {'A', 'B'} )
% Now, the following syntax returns a vector of doubles
X = ds.A;
whos X
% But this syntax returns a dataset
X = ds(:, 1);
whos X
% Which means this is valid:
ds.A = rand(10, 1) * 5;
% But this isn't:
ds(:, 1) = rand(10, 1) * 5;
% because you've got a dataset on the LHS but a vector of double on the RHS
% So to concat:
ds.B = strcat(ds.B, {' and '}, ds.B);
disp(ds)
