A0422
Title: On the use of random forest for two sample testing, with applications in empirical finance
Authors: Simon Hediger - University of Zurich (Switzerland) [presenting]
Jeffrey Naef - ETHZ (Switzerland)
Abstract: Several tests based on the Random Forest classifier are derived for the two-sample case. The tests are easy to use, require no tuning parameters, and are applicable for any $p$-dimensional distribution, notably for large $p$. The distribution of the test under the null of distributional equality is derived. Power analysis is conducted, both in theory, and with simulations. The R-package ``hypoRF'' contains the relevant codes. An application relevant to large-scale multivariate non-Gaussian GARCH models for financial asset returns is developed.
