Christophe Croux (K.U. Leuven)
Robust Sparse Principal Components Analysis
A method for principal component analysis is proposed that is sparse and robust at the same time. The sparsity delivers principal components that have loadings on a small number of variables, making them easier to interpret. The robustness makes the analysis resistant to outlying observations. The principal components correspond to directions that maximize a robust measure of the variance, with an additional penalty term to take sparseness into account. We propose an algorithm to compute the sparse and robust principal components. The method is applied on several real data examples, and diagnostic plots for detecting outliers and for selecting the degree of sparsity are provided. A simulation experiment studies the effect on statistical efficiency by requiring both robustness and sparsity.

Niels Haldrup (CREATES)
Discriminating between Fractional Noise and Spurious Long Memory
Many time series in economics and finance seem to exhibit long memory in the sense that the autocorrelations decline rather slowly. It is commonplace to model such persistence by fractionally integrated processes. However, it has been pointed out that many other types of processes that are indeed non-fractional can produce long memory. Examples includes time series processes with breaks and structural change as well as many non-linear time series models. It is of interest to distinguish between these different types of long memory processes since correct specification is essential for inference, model-building and forecasting. In this paper we exploit the fact that non-linear transformations of fractional noise processes will themselves exhibit long memory of a lower degree than the original series. In fact, the limiting degree of long memory is dictated by the Hermite rank of the corresponding non-linear transformation. The feature can be used as a device for discriminating fractional processes from other classes of long memory models.