Skip to main content
. Author manuscript; available in PMC: 2017 Dec 1.
Published in final edited form as: Psychol Methods. 2016 Dec;21(4):583–602. doi: 10.1037/met0000087

Algorithm 2.

Multivariate Tree Boosting with Covariance Discrepancy Loss

For m in 1, …, M steps (trees) do:

  1. For q in 1, …, Q outcome variables do:

    1. Fit tree m(q) to residuals, and compute the amount of covariance discrepancy Dm,q(4)

  2. Choose the outcome q* corresponding to the tree that produced the maximum covariance discrepancy Dm,q(4)

  3. Update residuals by subtracting the predictions of the tree fit to outcome q*, multiplied by step-size.