Mi. Griep et al., COMPARISON OF SEMIROBUST AND ROBUST PARTIAL LEAST-SQUARES PROCEDURES, Chemometrics and intelligent laboratory systems, 29(1), 1995, pp. 37-50
Three different robust partial least squares (PLS) alternatives are co
mpared in order to reduce the effect of outlying data. Although iterat
ively reweighted least squares, based on the biweight mean, offers eff
iciency in a variety of situations, in higher dimensions, there are st
ill considerable deviations from the ordinary PLS solution without out
liers.