This work is concerned with addressing the issue of variable selection in high-dimensional distributional regression models employed for causal inference. We regularise a Two-Stage Generalised Additive Model for Location, Scale, and Shape (2SGAMLSS) using component-wise gradient boosting in order to obtain a sparse model to assess the causal effect of rural electrification on female and male employment rates using socio-demographic data from South Africa.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados