Sounds like a job for categorical arrays! Huzzah! (Assuming you have Statistics Toolbox.) BTW, you said "dataset" but you're using cell arrays, so I assume you don't mean the dataset array in Stats TB. However, they may be a useful way to package your data. Anyway... why not make a new variable that is the combination of columns 2 and 3, and look for the unique values of that array:
twoandthree = nominal(strcat(data{2},'-',data{3}))
data = [data{:}];
domains = getlabels(twoandthree)
for k=1:length(domains)
foo = data(twoandthree==domains{k},:)
end
If you don't have Stats TB, you can achieve the same result with unique and strcmp:
twoandthree = strcat(data{2},'-',data{3})
data = [data{:}];
domains = unique(twoandthree)
for k=1:length(domains)
foo = data(strcmp(twoandthree,domains{k}),:)
end
Also, note I'm using [data{:}] to extract the four columns (each being a cell array) and concatenate them together into a single four-column table (ie a single n-by-4 cell array containing strings). If you're going to be accessing by rows, that's a nicer arrangement of data.
But, as I mentioned, dataset arrays may also make life nice, depending on what you're doing to do with the subsets.
data = dataset(data{:},'VarNames',strcat('cl',cellstr(num2str((1:4)'))))
twoandthree = nominal(strcat(data.cl2,'-',data.cl3))
domains = getlabels(twoandthree)
for k=1:length(domains)
foo = data(twoandthree==domains{k},:)
end
