What is 'cat' param used for in TreeBagger method

Question

dennis vella 2013-3-16

I am following the tutorial and am trying to implement TreeBagger Method. I have a question since I cannot understand part of the code.

b = TreeBagger(nTrees,X,Y,'oobpred','on','cat',6,'minleaf',leaf(ii));

can anyone tell me what 'cat' is and the number 6 please?

Thanks

Answer 1

Walter Roberson 2013-3-16

TreeBagger also accepts classregtree options.

categorical — Vector of indices of the columns of X that are to be treated as unordered categorical variables

dennis vella 2013-3-16

Does this mean that the number listed near 'cat' is possibly the number of columns that a dataset would have?

Answer 2

Tom Lane 2013-3-16

This means that column 6 of X is to be treated as categorical. Suppose it takes values 1,2,3. If X were continuous, splits would be of the form

    x<1.5    % separates 1 from {2,3}, or
    x<2.5    % separates {1,2} from 3

If X were categorical, then there is no ordering assumed so you could have a split that separates class 2 from classes {1,3}.