Research work · 2005
Partial Least Squares Regression
Inge Svein Helland
Encyclopedia of Statistical Sciences
Abstract
Abstract Partial least squares regression was introduced as an algorithm in the early 1980s, and it has gained much popularity in chemometrics. PLSR—or PLSR1—is a regression method for collinear data, and can be seen as a competitor to principal component regression. The method makes it possible to combine prediction with a study of a joint latent structure in thex,y‐variables. PLSR2 is a generalization to several dependent variables. PLSR1 can be understood from a statistical point of view via its parameter algorithm and the corresponding population model.