Identity spline variations in boosted Partial Least-Squares: a study on poverty
摘要
The aim of this paper is to propose a new functional way of practically modifying the observations of a statistical variable. It mainly draws its applications from the regression context to experiment with desirable changes in the response variable. We focus on some flexible variations around the identity spline function thus allowing to control local or global changes in the observations. Available for any regression model, this way of modifying a response is applied here in linear and nonlinear Partial Least-Squares models to explore how to alleviate poverty in a social context.