2016
DOI: 10.1016/j.sab.2016.07.010
|View full text |Cite
|
Sign up to set email alerts
|

Comparison of univariate and multivariate models for prediction of major and minor elements from laser-induced breakdown spectra with and without masking

Abstract: This study uses 1356 spectra from 452 geologically-diverse samples, the largest suite of LIBS rock spectra ever assembled, to compare the accuracy of elemental predictions in models that use only spectral regions thought to contain peaks arising from the element of interest versus those that use information in the entire spectrum. Results show that for the elements Si,

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
8
0

Year Published

2017
2017
2025
2025

Publication Types

Select...
6
2

Relationship

1
7

Authors

Journals

citations
Cited by 44 publications
(8 citation statements)
references
References 32 publications
0
8
0
Order By: Relevance
“…Stipe et al reported an error of ±8% at 6100 ppm K 2 O using an ICCD camera as a detector. Furthermore, measuring the full range of elemental abundances and using the partial least squares (PLS) regression could improve the accuracy of K 2 O abundances . The use of sub‐model training sets, which are composed of calibration samples having a more relevant concentration range, was shown to improve measurement capabilities .…”
Section: Age Measurement Results and Discussionmentioning
confidence: 99%
“…Stipe et al reported an error of ±8% at 6100 ppm K 2 O using an ICCD camera as a detector. Furthermore, measuring the full range of elemental abundances and using the partial least squares (PLS) regression could improve the accuracy of K 2 O abundances . The use of sub‐model training sets, which are composed of calibration samples having a more relevant concentration range, was shown to improve measurement capabilities .…”
Section: Age Measurement Results and Discussionmentioning
confidence: 99%
“…Univariate models have the advantage of being simple and easy to interpret, and the user has complete control over which emission line is used. However, although univariate models can be effective for minor and trace elements with few weak emission lines (e.g., [31,36]), they tend not to perform as well as multivariate models for major elements [37,66]. Multivariate models can incorporate information from the entire spectrum, including emission from elements other than the element being predicted, to mitigate matrix effects outlined above.…”
Section: Multivariate Vs Univariate Calibrationmentioning
confidence: 99%
“…tuning the parameters such that the model performs well on the training set but does not handle novel data well) or because the new observation represents an unusual composition relative to the samples on which the model was trained. If trained on a large and representative data set, the multivariate models have superior performance [37,38]; we therefore chose to focus our efforts on the multivariate approach for the major element calibrations discussed in this paper.…”
Section: Multivariate Vs Univariate Calibrationmentioning
confidence: 99%
“…There is no doubt that the multivariate analysis is superior to univariate analysis. This has been proven by numerous researchers, for instance, the analysis of rocks [49], rare earth elements [38], glass [50], cerium oxide [51], alloy steel [52], liquid steel [53], soil [54,55], soybean oil [56], PZT (Lead Zirconate Titanate) ceramics [57], Pb in navel orange [58], Marcellus Shale [59], tailing cores [60], geologically diverse samples [49], steel melt [61], slurry [62], iron ore [63] and pellets of plant materials [64].…”
Section: The Comparison Of Calibration Methodsmentioning
confidence: 81%