How to partition a table or matrix of observation data so that samples from every group are included?

I have a table of categorized data with 41 categories. Every category has at least 5 samples. If I partition it into a 90:10 training:test split like so:

cvp = cvpartition(data.Category,'Holdout',0.1);

it'll split it into the quantities I want but one group will frequently not contain samples from all 41 categories. How do I split it so that all categories are represented in both the training and test datasets? It's in Table format at the moment but can switch to cell array if needed.

Answer 1

[Sachin Meena]

Try the

cvpartition(group,'HoldOut',p,'Stratify',stratifyOption)

option, refer to cvpartition documentation for examples. You may not need to use tall arrays.