Integrating Single Index Effects in Generalized Additive Models

Linearly combining the elements of a vector of covariates to get a scalar-valued feature is common practice in regression modelling. In this work, we propose a novel approach to integrate single index effects in Generalised Additive Models (GAMs). In particular, model fitting and inference are performed by exploiting the efficient methods proposed in [7]. We consider an application to daily electricity load consumption data, demonstrating improved predictive performance relative to traditional GAMs. This integrated approach provides a valuable tool to capture complex relationships in real-world applications, while preserving interpretability.