Title: Clustered covariate regression
Authors: Emmanuel Tsyawo - Universite Mohammed VI Polytechnique (Morocco) [presenting]
Abdul-Nasah Soale - Case Western Reserve University (United States)
Abstract: High dimensionality is an increasingly occurrent phenomenon in model estimation. A common approach to handling high-dimensionality is regularisation-based methods that impose sparsity, and require that several elements of the high-dimensional parameter be zero. However, sparsity cannot always be assumed or easily verified in given empirical contexts. Severe bias and misleading inference may occur when sparsity does not hold. The Grouped Parameter Estimator is introduced, which generalises the notion of sparsity. It remains valid even if the support of the high-dimensional parameter is bounded away from zero. Monte Carlo simulations demonstrate the estimator's high approximative ability of the high-dimensional parameter, improved precision, reduced bias, and a favourably competitive performance relative to competing estimators.