Nonlinear Iterative Partial Least Squares (NIPALS) Modelling: Some Current Developments

Abstract

This chapter focuses on the potential use of the NIPALS approach in the recent many-faceted development of path models in the social sciences. Three general types of models are considered, denoted L, L*, and L**, and are distinguished by which categories of the explanatory variables are directly or indirectly observed. The models have a background in two central lines of evolution in econometrics and the behavioral sciences. In another perspective of multivariate analysis, the models belong under two broad lines of evolution—from high degrees of information in the modelling to lower degrees of information. In the social sciences in the past, there has been a forceful development of causal model building by the use of path models. A new and most significant feature is that econometric methods have been incorporated in the modelling, a merger that brings together widely different lines of evolution and thereby opens up new vistas and broadens the scope of the approach. The chapter explores these developments from the point of view of NIPALS modelling. The chapter reviews causal flow models in econometrics and the behavioral sciences.