Regression model checking with Berkson measurement errors

Koul, Hira L.; Song, Weixing

doi:10.1016/j.jspi.2007.05.048

Cited by 19 publications

(8 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this regard, one alternative approach could have been the development of "errors-in-variables" models that explicitly accommodate the uncertainty of the predictor variables, assuming that complete characterization of the associated error is feasible. 26 …”

Section: ■ Materials and Methodsmentioning

confidence: 99%

Evaluation and Interconversion of Various Indicator PCB Schemes for ∑PCB and Dioxin-Like PCB Toxic Equivalent Levels in Fish

Gandhi

Bhavsar

Reiner

et al. 2014

Environ. Sci. Technol.

View full text Add to dashboard Cite

Polychlorinated biphenyls (PCBs) remain chemicals of concern more than three decades after the ban on their production. Technical mixture-based total PCB measurements are unreliable due to weathering and degradation, while detailed full congener specific measurements can be time-consuming and costly for large studies. Measurements using a subset of indicator PCBs (iPCBs) have been considered appropriate; however, inclusion of different PCB congeners in various iPCB schemes makes it challenging to readily compare data. Here, using an extensive data set, we examine the performance of existing iPCB3 (PCB 138, 153, and 180), iPCB6 (iPCB3 plus 28, 52, and 101) and iPCB7 (iPCB6 plus 118) schemes, and new iPCB schemes in estimating total of PCB congeners (∑PCB) and dioxin-like PCB toxic equivalent (dlPCB-TEQ) concentrations in sport fish fillets and the whole body of juvenile fish. The coefficients of determination (R(2)) for regressions conducted using logarithmically transformed data suggest that inclusion of an increased number of PCBs in an iPCB improves relationship with ∑PCB but not dlPCB-TEQs. Overall, novel iPCB3 (PCB 95, 118, and 153), iPCB4 (iPCB3 plus 138) and iPCB5 (iPCB4 plus 110) presented in this study and existing iPCB6 and iPCB7 are the most optimal indicators, while the current iPCB3 should be avoided. Measurement of ∑PCB based on a more detailed analysis (50+ congeners) is also overall a good approach for assessing PCB contamination and to track PCB origin in fish. Relationships among the existing and new iPCB schemes have been presented to facilitate their interconversion. The iPCB6 equiv levels for the 6.5 and 10 pg/g benchmarks of dlPCB-TEQ05 are about 50 and 120 ng/g ww, respectively, which are lower than the corresponding iPCB6 limits of 125 and 300 ng/g ww set by the European Union.

show abstract

Section: ■ Materials and Methodsmentioning

confidence: 99%

Evaluation and Interconversion of Various Indicator PCB Schemes for ∑PCB and Dioxin-Like PCB Toxic Equivalent Levels in Fish

Gandhi

Bhavsar

Reiner

et al. 2014

Environ. Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…As a consequence, it is hard to determine the critical values used for testing. The same phenomena occur in the lack-of-fit test in classical regression models and measurement error models, see STZ, Koul and Song (2008) and the references therein.…”

Section: Test Statisticmentioning

confidence: 61%

“…To construct a distribution-free test statistic, STZ, Koul and Song (2008) applied the so-called Khamaladze type transformation on the test statistics. This transformation was first considered by Khamaladze (1981Khamaladze ( ,1988, and soon became a powerful tool for constructing distribution-free test statistics.…”

Section: Test Statisticmentioning

confidence: 99%

A lack-of-fit test in Tobit errors-in-variables regression models

Song

Yao

2011

Statistics & Probability Letters

Self Cite

View full text Add to dashboard Cite

“…The MD estimatorsθ n have the same property. Koul and Song [16] discussed a similar question in the regression model with Berkson measurement error, but their argument also applies to the current set-up. So, without loss of generality, we assume now that the estimatorsβ n and the MD estimatorθ n used in the test statistic satisfy…”

Section: Consistencymentioning

confidence: 93%

Minimum distance conditional variance function checking in heteroscedastic regression models

Samarakoon

Song

2011

Journal of Multivariate Analysis

Self Cite

View full text Add to dashboard Cite

a b s t r a c tThis paper discusses a class of minimum distance tests for fitting a parametric variance function in heteroscedastic regression models. These tests are based on certain minimized L 2 distances between a nonparametric variance function estimator and the parametric variance function estimator. The paper establishes the asymptotic normality of the proposed test statistics and that of the corresponding minimum distance estimator under the fitted model. These estimators turn out to be √ n-consistent. Consistency of this sequence of tests at some fixed alternatives and asymptotic power under some local nonparametric alternatives are also discussed. Some simulation studies are conducted to assess the finite sample performance of the proposed test.Published by Elsevier Inc.

show abstract

Regression model checking with Berkson measurement errors

Cited by 19 publications

References 15 publications

Evaluation and Interconversion of Various Indicator PCB Schemes for ∑PCB and Dioxin-Like PCB Toxic Equivalent Levels in Fish

Evaluation and Interconversion of Various Indicator PCB Schemes for ∑PCB and Dioxin-Like PCB Toxic Equivalent Levels in Fish

A lack-of-fit test in Tobit errors-in-variables regression models

Minimum distance conditional variance function checking in heteroscedastic regression models

Contact Info

Product

Resources

About