Weighting schemes for updating regression models—a theoretical approach

Stork, Chris L.; Kowalski, Bruce R.

doi:10.1016/s0169-7439(99)00016-7

Cited by 62 publications

(40 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Selection of a weight value in past work has been based on replication of samples in the standardization set [11,17]. For example, if l ¼ 1, then no replication of the standardization set is used, if l ¼ 2 then duplicates are augmented, etc.…”

Section: Tikhonov Regularization Modificationsmentioning

confidence: 99%

Impact of standardization sample design on Tikhonov regularization variants for spectroscopic calibration maintenance and transfer

Kunz

Ottaway

Kalivas

et al. 2010

Journal of Chemometrics

View full text Add to dashboard Cite

Multivariate spectroscopic calibration models are only valid to predict samples within the span of the calibration sample space measured relative to the current instrument environment (the primary conditions). Predicting samples in secondary conditions with new variances, same sample variances as the calibration space but a new instrument environment, or both, requires some form of continual model maintenance and/or transfer. Previous work has shown that a Tikhonov regularization (TR) approach is capable of accomplishing both tasks by updating the primary model based on only a few samples (transfer or standardization set) measured under the secondary conditions. A distinction of the TR design for calibration maintenance and transfer is a defined weighting scheme for the small set of standardization samples augmented to the full set of primary calibration samples. Critical to successful calibration maintenance or transfer is the standardization sample set composition, i.e. standardization samples should properly represent that are less secondary conditions. This paper reports on using TR-based methods to investigate this issue and a consensus modeling approach is briefly evaluated.

show abstract

Section: Tikhonov Regularization Modificationsmentioning

confidence: 99%

Impact of standardization sample design on Tikhonov regularization variants for spectroscopic calibration maintenance and transfer

Kunz

Ottaway

Kalivas

et al. 2010

Journal of Chemometrics

View full text Add to dashboard Cite

show abstract

“…When a calibration model developed from instrumental measurements of standards is to be used on an instrument differing from the one where the data were collected, it is not surprising to see substantially poorer predictive performance on new data collected over the range of calibration because the calibration model does not span the environmental and instrumental variations present in measurements made on the secondary instrument. Model updating (MUP), often used to improve calibrations on a single instrument to reflect a changing distribution of calibration samples [9,10], can be used to correct for the new instrumental variation by adding calibration samples taken on the secondary instrument to the calibration model [7,14], so that the updated calibration model now includes embedded instrumental background effects observed on both instruments. MUP of this type yields a new, global model that can be applied to either instrument, but because it now spans the additional instrumental variation, the updated calibration model is often more complex, with a larger number of latent variables in the final model.…”

Section: Transfer Of Calibration Using Model Updating and Orthogonal mentioning

confidence: 99%

“…Previous work has shown that, for similar instruments and for small calibration sets, a portion of the original calibration used as update samples, comprising 25-33% of the original calibration selected using a Kennard-Stone design and measured on the secondary instrument, is sufficient to obtain a global calibration model with performance similar to that of the original, single-instrument calibration model [7,14,15]. The update standards need not match those used in the original calibration, but with a large calibration set, it is necessary to either have many update standards or to weight them so that the update samples contribute sufficiently to the global model [9,10].…”

Section: Transfer Of Calibration Using Model Updating and Orthogonal mentioning

confidence: 99%

“…Use of a calibration on data from a secondary instrument that fall in the range of the calibration can sometimes be done by an update of the spectral calibration model to account for differences in the instrumental transfer function or in the environmental contributions to spectra. It is possible to update the calibration model for these differences directly [9,10], by adding spectra of a set of standards taken on the secondary instrument to the calibration set to provide a source of information on any changes in the instrumental transfer function or changes in the environmental contributions to the spectral signals and regenerating the calibration model. This new model is global to spectra obtained on both primary and secondary instruments.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Stacked PLS for calibration transfer without standards

Brown

Man

2011

Journal of Chemometrics

View full text Add to dashboard Cite

aWe report the successful application of stacked partial least-squares (SPLS) regression for direct application of multivariate calibration models to data from a secondary spectrometer, without use of any calibration transfer. Unlike a conventional calibration that requires transfer methods which need measurement of a set of transfer samples to make useful predictions from data obtained on a secondary instrument, SPLS regression can be used to generate regression models with good predictive power on both primary and secondary instruments. Results of direct application of SPLS to lake sediment data without use of any transfer standards show predictive results comparable to those obtained from conventional PLS calibration followed by a model updating (MUP) step. We also demonstrate the use of calibration MUP applied to stacked PLS regression, which further minimizes local differences between two instruments.

show abstract

“…Another approach [10] consists of developing a −generic model× that models the intraobject signal effect separately and that can be adapted to a target object. Also it is usual to update a calibration model by adding new samples to the calibration set in order to improve the model [11], the weak point of this approach being to define how −similar× a new sample is to the samples currently defining the calibration set [12]. Calibration strategies have even been developed based on locally weighted regressions with extended libraries [13,14].…”

Section: Introductionmentioning

confidence: 99%

Maintenance of Soft Calibration Models in the Determination of Zinc, Cadmium, Lead and Copper by Differential Pulse Anodic Stripping Voltammetry

et al. 2004

View full text Add to dashboard Cite

A strategy for constructing a global multivariate calibration model that includes calibration samples measured over time on different days is developed and applied in electroanalysis. Both synthetic and real samples (tap, extracted and river water) are analyzed by differential-pulse anodic stripping voltammetry, showing the suitability of the global model constructed that provides successful results similar to those of the usual multivariate calibration. In addition the capability of discrimination of this model is evaluated in prediction for the mean of three replicates with estimation of probability of false noncompliance, a, and false compliance, b, being found 3.1, 11.2, 6.7 and 64.7 nM for nominal concentrations of zinc, cadmium, lead and copper of 96.0, 40.4, 37.3 and 328.0 nM respectively when a b 0.05. It has been proven that the use of the global calibration does not imply a loss of multivariate analytical sensitivity, using this parameter as quality index of the analytical procedure. The viability of using calibration maintenance strategies with electroanalytical techniques is shown, providing a way to save time and experimental effort when these techniques are used in routine analysis.

show abstract

Weighting schemes for updating regression models—a theoretical approach

Cited by 62 publications

References 15 publications

Impact of standardization sample design on Tikhonov regularization variants for spectroscopic calibration maintenance and transfer

Impact of standardization sample design on Tikhonov regularization variants for spectroscopic calibration maintenance and transfer

Stacked PLS for calibration transfer without standards

Maintenance of Soft Calibration Models in the Determination of Zinc, Cadmium, Lead and Copper by Differential Pulse Anodic Stripping Voltammetry

Contact Info

Product

Resources

About