Independence tests for financial variables

This paper proposes an alternative method to evaluate the independence between random variables. The new method is particularly useful when the tested random variables are continuous, because the most used tests for independence are not able to give precise evaluations. In particular, we analyze and compare two different methods to test the independence among financial variables. The first is the classical chi-squared test generally used to evaluatethe independence of historical observations in the portfolio risk valuation. The new alternative method is based on a conditional expectation estimator. Thus, we can compare the results of the two methods by evaluating the performance in terms of goodness-of-fittests.