Variable screening based on Gaussian Centered L-moments

An important challenge in big data is identification of variables. For this purpose, methods discovering variables with non-standard univariate marginal distributions are proposed. The conventional moments based summary statistics can be well-adopted, but their sensitivity to outliers lead selection on a few rather than distributional shape such as bimodality. To address type non-robustness, the L-moments considered. Using these practice, however, has limitation since they do not take zero values at Gaussian which distribution most naturally compared. As remedy, Centered proposed, share advantages have zeros distributions. strength over other shown theoretical and practical aspects performances screening genes cancer genetics data.