International Conference on Robust Statistics (ICORS) location:Halle (Germany) date:10-15 August 2014
Canonical correlation analysis (CCA) describes the associations between two sets of variables by maximizing the correlation between linear combinations of the variables in each data set. This talk discusses a method for Robust Sparse CCA. Sparse estimation produces linear combinations of only a subset of variables from each data set. More precisely, some of the elements of the canonical vectors will be estimated as exactly zero. As such, the interpretability of the canonical variates is increased. We also robustify the method such that it can cope with outliers in the data. To estimate the canonical vectors, we convert the CCA problem into an alternating regression framework. Sparse canonical vectors are obtained by adding a lasso penalty on the coefficient estimates to the Least Squares estimator. The lasso, however, is not robust to outliers. The method can be easily robustified by using the sparse Least Trimmed Squares estimator. We illustrate the good performance of the Robust Sparse CCA method in several simulation studies.