Regression by local fitting: Methods, properties, and computational algorithms

doi:10.1016/0304-4076(88)90077-2

Journal of Econometrics

Volume 37, Issue 1, January 1988, Pages 87-114

https://doi.org/10.1016/0304-4076(88)90077-2 Get rights and content

Abstract

Local regression is a procedure for estimating regression surfaces by the local fitting of linear or quadratic functions of the independent variables in a moving fashion that is analogous to how a moving average is computed for a time series. The advantage of the methodology over the global fitting of parametric functions of the independent variables by least squares, the current paradigm in regression studies, is that a much wider class of regression functions can be estimated without distortion. In this paper, we discuss the methods, their statistical properties, and computational algorithms.

References (38)

R.E. Barnhill
Representation and approximation of surfaces
C. de Boor et al.
An adaptive algorithm for multivariate approximation giving optimal convergence rates
Journal of Approximation Theory
(1979)
R.E. Bank et al.
Refinement algorithms and data structures for regular local mesh refinement
D.A. Belsley et al.
Regression diagnostics
(1980)
P. Bloomfield
Fourier analysis of time series: An introduction
(1976)
L. Breiman et al.
Estimating optical transformations for correlation and regression
Journal of the American Statistical Association
(1985)
J.M. Chambers et al.
Graphical methods for data analysis
(1983)
W.S. Cleveland
Robust locally-weighted regression and smoothing scatterplots
Journal of the American Statistical Association
(1979)
W.S. Cleveland et al.
Statistical properties of locally-weighted regression: Field and laboratory simulations
Bell Communications Research technical memorandum and AT&T Bell Laboratories technical memorandum
(1987)
Cleveland, W.S. and S.J. Devlin, to appear, Locally-weighted regression: An approach to regression analysis by local...

G. Collomb

Estimation non-paramétrique de la regression: Revue bibliographique

International Statistical Review

(1981)

R.D. Cook et al.

Influence and residuals in regression

(1982)

P. Craven et al.

Smoothing noisy data with spline functions: Estimating the correct degree of smoothing by the method of generalized cross-validation

Numerische Mathematik

(1979)

N. Cressie

Kriging nonstationary data

Journal of the American Statistical Association

(1986)

S.J. Devlin

Locally-weighted multiple regression: Statistical properties and its use to test for linearity

Bell Communications Research technical memorandum

(1986)

A.S.C. Ehrenberg

A primer in data reduction

(1982)

B. Efron

The jackknife, the bootstrap, and other resampling plans

(1982)

R.W. Floyd et al.

Algorithm 489: The algorithm SELECT — for finding the i-th smallest of n elements

Communications of the ACM

(1975)

E.B. Fowlkes

Some diagnostics for binary logistic regression via smoothing

Bell Communications Research technical memorandum

(1986)

Cited by (447)

SCORE2 Report 24: Nonlinear Relationship of Retinal Thickness and Visual Acuity in Central Retinal and Hemiretinal Vein Occlusion
2023, Ophthalmology
To investigate whether a nonlinear association between central subfield thickness (CST) on spectral-domain OCT and concurrent visual acuity letter score (VALS) exists in eyes treated initially with aflibercept or bevacizumab for macular edema associated with central retinal vein occlusion (CRVO) or hemiretinal vein occlusion (HRVO) in the Study of Comparative Treatments for Retinal Vein Occlusion 2 (SCORE2).
Long-term follow-up after a randomized clinical trial from 64 centers in the United States.
Participants were followed up to 60 months and treated at investigator discretion after completing the 12-month treatment protocol.
Two-segment linear regression models were compared with simple linear regression models of VALS on CST. Pearson correlation coefficients were calculated to assess strength of CST and VALS associations.
Central subfield thickness was measured by OCT and VALS by the electronic Early Treatment Diabetic Retinopathy Study methodology.
Estimated inflection points, reflecting turning points at which the CST and VALS association changes from positive to negative, calculated at 7 postbaseline visits, range from 217 to 256 μm. A strongly positive correlation exists to the left of each estimated inflection point, ranging from 0.29 (P < 0.01 at month 60) to 0.50 (P < 0.01 at month 12), and a strongly negative correlation exists to the right of each estimated inflection point, ranging from −0.43 (P < 0.01 at month 1) to −0.74 (P < 0.01 at month 24). Randomization statistical tests showed that 2-segment models are favored over 1-segment models for all postbaseline months (P < 0.001 for all tests performed).
The relationship between CST and VALS in eyes with CRVO or HRVO after treatment with anti–vascular endothelial growth factor (VEGF) therapy is not simply linear. The usually modest correlations between OCT-measured CST and visual acuity belie strong left and right correlations present in 2-segment models. Post-treatment CST close to the estimated inflection points showed the best expected VALS. The SCORE2 participants with a post-treatment CST after treatment close to the estimated inflection points of 217 to 256 μm showed the best VALS. In patients treated with anti-VEGF for macular edema associated with CRVO or HRVO, a thinner retina is not always associated with better VALS.
Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.
Risk of prolonged ischemic time linked to use of cardiopulmonary bypass during implantation for lung transplantation in the United Kingdom
2023, Journal of Heart and Lung Transplantation
Some degree of ischemia is inevitable in organ transplantation, and for most, if not all organs, there is a relationship between ischemic time and transplant outcome. The contribution of ischemic time to lung injury is unclear, with conflicting recent data. In this study, we investigate the impact of ischemia time on survival after lung transplantation in a large national cohort.
We studied the outcomes for 1,565 UK adult lung transplants over a 12-year period, for whom donor, transplant, and recipient data were available from the UK Transplant Registry. We examined the effect of ischemia time (defined as donor cross-clamp to recipient reperfusion) and whether standard cardiopulmonary bypass was used using Cox proportional hazards models, adjusting for other risk factors.
The total ischemic time increased from a median under 5 hours in 2003 to over 6.2 hours in 2013. Our findings show that, when the cardiopulmonary bypass was used, there was an increase in the hazard of death (of 13% [95% CI: 5%-21%] for 1-year patient survival) for each hour of total ischemic time. However, if the cardiopulmonary bypass was not used for implantation, this link disappeared—there was no statistically significant change in mortality with increasing ischemic time.
We document that avoidance of bypass may remove ischemic time, within the limits of our observed range of ischemic times, as a risk factor for poor outcomes. Our data add to the evidence that bypass may be harmful to the donor lung.
Time series, seasonality and trend evaluation of 7 years (2015–2021) of OSHA severe injury data
2023, Journal of Safety Research
Problem: Employers are required to report severe work-related injuries (e.g., amputation, inpatient hospitalization, or loss of an eye), to the Occupational Safety and Health Administration (OSHA). This study examined the OSHA severe injury reports (SIRs) public microdata to understand time-related trends and patterns. Methods: This study included all SIRs from January 2015 to December 2021 (84 months). We employed time series decomposition models (classical additive and multiplicative, X-11, and X-13ARIMA-SEATS) to evaluate monthly seasonal effect and seasonally adjusted trend of SIRs. We developed data visuals to display trends from different models with the original data series. We compared number of daily SIRs by day of the week, and yearly trends by 2-digit NAICS and separately by 1-digit OIICS injury event. Results: There were a total of 70,241 SIRs in this 7 year period; ranging from 8,704 to 11,156 per year, and 600 to 1,100 per month. Seasonally adjusted trend indicated a gradual increase of SIRs over time until October 2018, then a steeper decrease until August 2020, and staying somewhat flat for the rest of the months. Seasonality indicated more SIRs were reported in the summer months (June, July, August). Daily SIRs indicated a weekday average of 34 (SD = 9) and weekend average of 11 (SD = 5). The Manufacturing and Construction industries reported the highest yearly SIRs. Contact with objects and equipment, and falls, slips, trips were the most numerous injury events associated with SIRs. Discussion: Although Federal OSHA SIR data do not include SIRs from state-plan jurisdictions, the data provide a timely national trend of SIR. This is the first known time series analysis of SIRs. Practical Applications: The findings of this study highlight the ability of researchers to use the SIRs as a timely indicator to understand occupational injury trends by specific industries and injury events.
Ensemble habitat suitability modeling for predicting optimal sites for eelgrass (Zostera marina) in the tidal lagoon ecosystem: Implications for restoration and conservation
2023, Journal of Environmental Management
Seagrass systems are in decline, mainly due to anthropogenic pressures and ongoing climate change. Implementing seagrass protection and restoration measures requires accurate assessment of suitable habitats. Commonly, such assessments have been performed using single-algorithm habitat suitability models, nearly always based on low environmental resolution information and short-term species data series. Here we address eelgrass (Zoostera marina) meadows’ large-scale decline (>80%) in Shandong province (Yellow Sea, China) by developing an ensemble habitat model (EHM) to inform eelgrass conservation and restoration strategies in the Swan Lake (SL). For this, we applied a weighted EHM derived from ten single-algorithm models including profile, regression, classification, and machine learning methods to generate a high-resolution habitat suitability map. The EHM was constructed based on the predictive performances of each model, by combining a series of present-absent eelgrass datasets from recent years coupled with oceanographic and sediment data. The model was cross-validated with independent historical datasets, and a final habitat suitability map for conservation and restoration was generated. Our EHM scheme outperformed all single models in terms of habitat suitability, scoring ∼0.95 for both true statistic skill (TSS) and area under the curve (AUC) performance criteria. Machine learning methods outperformed profile, regression and classification methods. Regarding model explanatory variables, overall, topographic characteristics such as depth (DEP) and seafloor slope (SSL) are the most significant factors determining the distribution of eelgrass. The EHM predicted that the overlapping area was almost 90% of the current eelgrass habitat. Using results from our EHM, a LOESS regression model for the relationship of the habitat suitability to both the biomass and density of Z. marina outperformed better than the classic Ordinary Least Squares regression model. The EHM is a promising tool for supporting eelgrass protection and restoration areas in temperate lagoons as data availability improves.
A thorough comparison between measurements and predictions of the amplitude dependent natural frequencies and damping of a bolted structure
2023, Journal of Sound and Vibration
Bolted joints are a significant source of damping and nonlinearity in built up structures. In the micro-slip regime (i.e. when the bolts do not slip completely) the log of the damping tends to increase linearly with the log of the vibration amplitude, the so-called power-law behavior. The natural frequency also tends to decrease slightly with vibration amplitude. While many works have successfully tuned phenomenological models to capture these behaviors, far fewer have sought to predict the nonlinearity in a bolted joint and validated the predictions with measurements over a range of bolt preloads and vibration amplitudes. This work presents a step towards such a prediction, using a detailed finite element model of a structure, including the preload forces in the bolts and Coulomb friction in the contact, to seek to predict the nonlinear damping and stiffness of the structure and how they vary with preload. The structure studied consists of two beams bolted together at their free ends, the so-called S4 Beam. While the contact interfaces were machined to be nominally flat, the micron-scale curvature in the contact interface is included in the model, approximated in two different ways, to seek to understand what fidelity is needed at the contact interface to successfully capture its dynamic behavior. The recently presented Quasi-Static Modal Analysis (QSMA) method is employed, where the amplitude dependent damping and natural frequency can be predicted from a single quasi-static simulation, avoiding the expense of numerical integration. The predicted damping and natural frequency are compared with measurements at various preloads, showing reasonable agreement if a Coulomb friction coefficient of 0.2 is used for all simulations. The simulations also revealed that, while the micron-scale curvature of the interfaces was important, the results were not very sensitive to how it was approximated.
Diagnosis-to-ablation time predicts recurrent atrial fibrillation and rehospitalization following catheter ablation
2022, Heart Rhythm O2
Wait times for catheter ablation in patients with symptomatic atrial fibrillation (AF) may influence clinical outcomes.
This study examined the relationship between the duration from AF diagnosis to ablation, or diagnosis-to-ablation time (DAT), on the clinical response to catheter ablation in a large nationwide cohort of patients.
We identified patients with new AF who underwent catheter ablation between January 2014 and December 2017 using the IBM MarketScan databases. Cox proportional hazard models were used to estimate the strength of the association between DAT and the outcomes of AF recurrence and hospitalization at 1 year postablation.
Among 11,143 AF patients who underwent ablation, the median age was 59 years, 31% were female, and the median CHA₂DS₂-VASc score was 2. Median DAT was 5.5 (2.6, 13.1) months. At 1 year postablation, 10.0% (n = 1116) developed recurrent AF. For each year increase in DAT, the risk of AF recurrence increased by 20% after adjustment for baseline comorbidities and medications (hazard ratio [HR] 1.20, 95% confidence interval [CI] 1.11–1.30). A longer DAT was associated with an increased risk of hospitalization (HR 1.08 per DAT year, 95% CI 1.02–1.15). DAT was a stronger predictor of AF recurrence postablation than traditional clinical risk factors, including age, prior heart failure, or renal failure.
Increasing duration between AF diagnosis and catheter ablation is associated with higher AF recurrence rates and all-cause hospitalization. Our findings are consistent with a growing body of evidence supporting the benefits of prioritizing early restoration of sinus rhythm.

View all citing articles on Scopus

View full text

Regression by local fitting: Methods, properties, and computational algorithms

Abstract

Journal of Approximation Theory

Refinement algorithms and data structures for regular local mesh refinement

Regression diagnostics

Fourier analysis of time series: An introduction

Estimating optical transformations for correlation and regression

Journal of the American Statistical Association

Graphical methods for data analysis

Robust locally-weighted regression and smoothing scatterplots

Journal of the American Statistical Association

Statistical properties of locally-weighted regression: Field and laboratory simulations

Bell Communications Research technical memorandum and AT&T Bell Laboratories technical memorandum

Estimation non-paramétrique de la regression: Revue bibliographique

International Statistical Review

Influence and residuals in regression

Smoothing noisy data with spline functions: Estimating the correct degree of smoothing by the method of generalized cross-validation

Numerische Mathematik

Kriging nonstationary data

Journal of the American Statistical Association

Locally-weighted multiple regression: Statistical properties and its use to test for linearity

Bell Communications Research technical memorandum

A primer in data reduction

The jackknife, the bootstrap, and other resampling plans

Algorithm 489: The algorithm SELECT — for finding the i-th smallest of n elements

Communications of the ACM

Some diagnostics for binary logistic regression via smoothing

Bell Communications Research technical memorandum