How does 'InBagFraction' work in Treebagger?

I have created a TREEBAGGER object setting 'oobvarimp' to 'on'. I want to determine what fraction of observations are used as out-of-bag observations.

I know the default value of InBagFraction is one and that the training data set is two-thirds of the total. How can I set the training data set to 80%?

Answer 1

[Raunak Gupta]

Hi,

The ‘InBagFraction’ is a fraction of input data to be used while creating a new tree. It has nothing to do with amount of out-of-bag sample as mentioned here. Since maximum value for ‘InBagFraction’ can be 1 so it is not possible to get 80% of the input data as in-bag observations. From the trained Model you may see that Mdl.OOBIndices will return a matrix. The true (1) values in matrix represent out-of-bag observation for that tree. Counting all non-zero value in that matrix will return overall out-of-bag samples count and thus give you the fraction.