Pseudo-degrees of freedom for complex predictive models: the example of partial least squares

@article{citeulike:2601367, abstract = {This paper considers models that are relatively complex considering the extent of the calibration data. Such data frequently arise in chemometric applications, near-infrared spectroscopy (NIRS) being a well-known example. Commonly used models are multiple linear regression (MLR) with variable selection or partial least squares (PLS) regression. The concept of degrees of freedom is undefined for such models; this paper proposes a definition for pseudo-degrees of freedom (PDF) based on predictive performance and an analogy with the standard linear model. The generalization is intended for all models which assume independent and identically distributed errors. Pseudo-degrees of freedom are very easily calculated from ordinary and cross-validation residuals. An example from a real-life NIRS application is given to illustrate the new concept. Copyright {\copyright} 1999 John Wiley \& Sons, Ltd.}, address = {Centre for Biometry Wageningen (CBW), PO Box 16, NL-6700 AA Wageningen, The Netherlands}, author = {van der Voet, Hilko }, citeulike-article-id = {2601367}, doi = {http://dx.doi.org/10.1002/(SICI)1099-128X(199905/08)13:3/4\%3C195::AID-CEM540\%3E3.0.CO;2-L}, journal = {Journal of Chemometrics}, keywords = {cybernetics}, number = {3-4}, pages = {195--208}, posted-at = {2008-04-05 13:59:25}, priority = {2}, title = {Pseudo-degrees of freedom for complex predictive models: the example of partial least squares}, url = {http://dx.doi.org/10.1002/(SICI)1099-128X(199905/08)13:3/4\%3C195::AID-CEM540\%3E3.0.CO;2-L}, volume = {13}, year = {1999} }

TY - JOUR ID - citeulike:2601367 L3 - citeulike-article-id:2601367 TI - Pseudo-degrees of freedom for complex predictive models: the example of partial least squares JF - Journal of Chemometrics VL - 13 IS - 3-4 SP - 195 EP - 208 AD - Centre for Biometry Wageningen (CBW), PO Box 16, NL-6700 AA Wageningen, The Netherlands N2 - This paper considers models that are relatively complex considering the extent of the calibration data. Such data frequently arise in chemometric applications, near-infrared spectroscopy (NIRS) being a well-known example. Commonly used models are multiple linear regression (MLR) with variable selection or partial least squares (PLS) regression. The concept of degrees of freedom is undefined for such models; this paper proposes a definition for pseudo-degrees of freedom (PDF) based on predictive performance and an analogy with the standard linear model. The generalization is intended for all models which assume independent and identically distributed errors. Pseudo-degrees of freedom are very easily calculated from ordinary and cross-validation residuals. An example from a real-life NIRS application is given to illustrate the new concept. Copyright © 1999 John Wiley & Sons, Ltd. KW - cybernetics AU - van der Voet, Hilko PY - 1999/// UR - http://dx.doi.org/10.1002/(SICI)1099-128X(199905/08)13:3/4%3C195::AID-CEM540%3E3.0.CO;2-L ER -