Chemometrics with R by Ron Wehrens

Author:Ron Wehrens
Language: eng
Format: epub
ISBN: 9783662620274
Publisher: Springer Berlin Heidelberg

Although the selection of the optimal number of components is a subjective process and should be guided by background knowledge as well as common sense, there are situations where a more automatic approach is desirable, e.g., in large-scale simulation studies. The pls package provides the function selectNcomp, implementing two different strategies. The first and simplest one estimates standard deviations for the RMSEP estimates in crossvalidation error curves such as shown in Fig. 8.2, and picks the simplest model that is within one standard deviation of the global optimum in the curve. This heuristical approach has been suggested in the context of methods like ridge regression and the lasso (Hastie et al. 2001) but can also be applied here. The second approach implements a permutation test (see also Sect. 9.​3.​1), again using the global minimum in the crossvalidation curve as a reference, and testing whether simpler models are significantly worse (Van der Voet 1994). Both approaches also provide graphical output. The input is a fitted model, including validation data:

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.