Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “regression”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,027 records · Page 57
All sourcesNASA NTRSDOE OSTI
DOE OSTI2016

SU-F-T-118: Characterization of Change in Fractional Anisotropy After Radiation Therapy: Does Nearby Disruption Predict for White Matter Damage?

Purpose: We investigated predictors of fractional anisotropy (FA) change in the corticospinal white matter tract (CST) following radiation therapy (RT). Methods: Diffusion tensor imaging (DTI) is a non-invasive modality which models water diffusion properties. FA quantifies the extent of directional bias—a decrease indicates disrupted white matter integrity. Fifteen patients with high-grade glioma underwent DTI scans before, and ten months after RT to 59.4–60 Gy. The CST was segmented using an automated atlas-based algorithm on all DTI images. Treatment planning CT and DTI images were aligned using non-linear registration allowing for baseline FA, follow-up FA, and absorbed dose to be determined in each voxel. Relative FA change was dichotomized into a binary outcome using 25% decrease as cutoff. Three metrics were assessed as predictors: voxel dose, distance from the voxel to the center of the CST (Rc), and the number of neighboring voxels (Nadj from 0 to 26) with ≥25% decrease in FA. Logistic regression and the area under the receiver-operating characteristics curve (AUC) analysis were performed for each patient. Results: Median age of the cohort was 59 years (range: 40–85). The average number of voxels in the CST amongst all patients was 1181 (±172, SD). In logistic regression, the probability of FA change was highly associated with Nadj in all 15 patients with corresponding AUCs between 0.81 and 0.97. With all three metrics included in the logistic regression models, Nadj was highly significant (p<0.001) in all patients, voxel dose significant (p<0.05) in 3/15 patients, and Rc significant in 12/15 patients (p<0.05). Conclusion: The number of neighboring voxels with change in FA was the dominant predictor of FA change at any given voxel. This suggests that the microenvironment of surrounding white matter disruption after radiation therapy may drive local effects along a white matter tract. Pettersson and Cervino are funded by a Varian Medical Systems grant.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2016

SU-G-BRA-09: Estimation of Motion Tracking Uncertainty for Real-Time Adaptive Imaging

Purpose: kV fluoroscopic imaging combined with MV treatment beam imaging has been investigated for intrafractional motion monitoring and correction. It is, however, subject to additional kV imaging dose to normal tissue. To balance tracking accuracy and imaging dose, we previously proposed an adaptive imaging strategy to dynamically decide future imaging type and moments based on motion tracking uncertainty. kV imaging may be used continuously for maximal accuracy or only when the position uncertainty (probability of out of threshold) is high if a preset imaging dose limit is considered. In this work, we propose more accurate methods to estimate tracking uncertainty through analyzing acquired data in real-time. Methods: We simulated motion tracking process based on a previously developed imaging framework (MV + initial seconds of kV imaging) using real-time breathing data from 42 patients. Motion tracking errors for each time point were collected together with the time point’s corresponding features, such as tumor motion speed and 2D tracking error of previous time points, etc. We tested three methods for error uncertainty estimation based on the features: conditional probability distribution, logistic regression modeling, and support vector machine (SVM) classification to detect errors exceeding a threshold. Results: For conditional probability distribution, polynomial regressions on three features (previous tracking error, prediction quality, and cosine of the angle between the trajectory and the treatment beam) showed strong correlation with the variation (uncertainty) of the mean 3D tracking error and its standard deviation: R-square = 0.94 and 0.90, respectively. The logistic regression and SVM classification successfully identified about 95% of tracking errors exceeding 2.5mm threshold. Conclusion: The proposed methods can reliably estimate the motion tracking uncertainty in real-time, which can be used to guide adaptive additional imaging to confirm the tumor is within the margin or initialize motion compensation if it is out of the margin.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

Physical Interpretation of Early Battery Life Prediction Models

Early battery life prediction models are most useful for R&D if they help us understand the early changes in battery electrochemical response that correspond with long-term degradation and failure. Linear regression models such as Fused lasso and Partial Least Squares can fit coefficients directly to high-dimensional electrochemical data like capacity-voltage and ΔV–state-of-charge, i.e., Q(V) and ΔV(SOC) curves, learning coefficients that can be physically interpreted. We leverage the ISU-ILCC battery aging data set to learn high-dimensional coefficients for early battery life prediction from traditional slow-rate capacity check data, demonstrating learning on Q(V), d Q· d V −1 , and ΔV(SOC) curves. A thorough study on the dependence of coefficient values on train/test size and data preprocessing methods is made, demonstrating the reliability of high-dimensional regression approaches unless very small amounts of data are used for model training. For this data set, coefficients from Q(V) and d Q· d V −1 models highlight changes in electrode stoichiometry due to lithium loss, while ΔV(SOC) coefficients highlight changes in positive electrode diffusivity due to particle cracking as well as electrode stoichiometry shifts. By directly interpreting the coefficients of a regression model, we make physical insights into battery degradation mechanisms without requiring the assumptions of traditional battery data analysis methods.

25 ENERGY STORAGE↗
DOE OSTI2021

Quantile-Dependent Expressivity of Serum Uric Acid Concentrations

“Quantile-dependent expressivity” occurs when the effect size of a genetic variant depends upon whether the phenotype (e.g., serum uric acid) is high or low relative to its distribution. Analyses were performed to test whether serum uric acid heritability is quantile-specific and whether this could explain some reported gene-environment interactions. Methods. Serum uric acid concentrations were analyzed from 2151 sibships and 12,068 offspring-parent pairs from the Framingham Heart Study. Quantile-specific heritability from offspring-parent regression slopes (β OP , h 2 = 2β OP /(1 + r spouse )) and full-sib regression slopes (β FS ,h 2 = {(1 + 8r spouse β FS ) 0.5 - 1}/(2r spouse ) was robustly estimated by quantile regression with nonparametric significance assigned from 1000 bootstrap samples. Results. Quantile-specific h 2 (±SE) increased with increasing percentiles of the offspring’s sex- and age-adjusted uric acid distribution when estimated from β OP (P trend = 0.001): 0.34 ± 0.03 at the 10th, 0.36 ± 0.03 at the 25th, 0.41 ± 0.03 at the 50th, 0.46 ± 0.04 at the 75th, and 0.49 ± 0.05 at the 90th percentile and when estimated from β FS (P trend = 0.006). This is consistent with the larger genetic effect size of (1) the SLC2A9 rs11722228 polymorphism in gout patients vs. controls, (2) the ABCG2 rs2231142 polymorphism in men vs. women, (3) the SLC2A9 rs13113918 polymorphism in obese patients prior to bariatric surgery vs. two-year postsurgery following 29 kg weight loss, (4) the ABCG2 rs6855911 polymorphism in obese vs. nonobese women, and (5) the LRP2 rs2544390 polymorphism in heavier drinkers vs. abstainers. Quantile-dependent expressivity may also explain the larger genetic effect size of an SLC2A9/PKD2/ABCG2 haplotype for high vs. low intakes of alcohol, chicken, or processed meats. Conclusions. Heritability of serum uric acid concentrations is quantile-specific.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2020

SISSO (Sure-independence-screening sparsifying-operator regressor)

Implementation of the SISSO regression algorithm in MATLAB. The SISSO regression algorithm iteratively selects model features from candidates, converging even when the number of possible features is much greater than the number of available data points. Includes a test script that validates the algorithm by replicating the results using data and procedure from https://analytics-toolkit.nomad-coe.eu/hub/user-redirect/notebooks/tutorials/compressed_sensing.ipynb. Python code from the SISSO regressor in 'sisso.py', from the link above, was used as the basis for developing the MATLAB implementation. The SISSO regression algorithm is detailed by the original authors in R. Ouyang, S. Curtarolo, E. Ahmetcik et al., Phys. Rev. Mater. 2, 083802 (2018), R. Ouyang, E. Ahmetcik, C. Carbogno, M. Scheffler, and L. M. Ghiringhelli, J. Phys.: Mater. 2, 024002 (2019).

Gasper, Paul↗
DOE OSTI2022

BPP R package

The BPP package contains R code to perform Bayesian projection pursuit regression under various regression settings. Bayesian projection pursuit regression learns the functional relationship between inputs and outputs by considering a sum of functions of one dimensional projections of the inputs. This is useful as a general machine learning software.

Francom, Devin↗
DOE OSTI2022

pnnl/Chemometric_Toolbox

The PNNL Chemometric Toolbox is a software collection of common MATLAB scripts that implement core chemometric algorithms for regression analysis. The three core regression techniques within this toolbox are classical least squares (CLS), principal component regression (PCR), and partial least squares (PLS). This is demonstrated using a supplied dataset of infrared (FTIR) spectral data with their corresponding concentrations

Smith, Ian↗
DOE OSTI2024

SysCaps (Language Interfaces for Simulation Surrogates of Complex Systems) [SWR-24-97]

You've found the official code repository for the paper "SysCaps: Language Interfaces for Simulation Surrogates of Complex Systems," presented at the Foundation Models for Science: Progress, Opportunities, and Challenges workshop at NeurIPS 2024. Our paper conjectures that interfaces (both text templates as well as conversational) makes interacting with simulation surrogate models for complex systems more intuitive and accessible for both non-experts and experts. "System captions", or SysCaps, are text-based descriptions of systems based on information contained in simulation metadata. Our paper's goal is to train multimodal regression models that take text inputs (SysCaps) and timeseries inputs (exogenous system conditions such as hourly weather) and regress timeseries simulation outputs (e.g. hourly building energy consumption). The experiments in our paper with building and wind farm simulators, which can be reproduced using this codebase, aim to help us understand whether a) accurate regression in this setting is possible and b) if so, how well can we do it. Paper: https://arxiv.org/abs/2405.19653

Emami, Patrick↗
DOE OSTI2025

Decayheatml

This code is designed to predict and analyze the decay heat generated in molten salt reactors (MSRs) using a hybrid approach that combines machine learning and segmented polynomial fitting. The accurate prediction of decay heat is essential for reactor safety and the optimization of spent fuel storage. The code operates through several key components: 1) Data Architecture: It incorporates a modular data architecture that handles various MSR-specific operational parameters such as power density, humidity content, and air ingress. These parameters are sampled using Sobol sequences to ensure comprehensive coverage of operational uncertainties. 2) Machine Learning Framework: The code employs a diverse set of machine learning models, including polynomial regression, decision trees, random forests, gradient boosting, support vector regression, k-nearest neighbors, multi-layer perceptrons, and symbolic regression. These models are trained to predict decay heat over a wide temporal range, from immediate shutdown up to 10,000 years. 3) Region-Optimized Training: The temporal domain is divided into multiple regions, each modeled separately to capture distinct decay heat characteristics across different time scales. This approach significantly improves the accuracy and interpretability of predictions. 4) Segmented Polynomial Interpretation (SPI): The SPI method translates machine learning predictions into piecewise polynomial equations. These equations are physically interpretable and can be directly integrated into existing engineering workflows and safety analyses. 5) Front-End Interfaces: The code includes both a Jupyter notebook interface for research development and a Streamlit web application for operational deployment. These interfaces allow users to interactively explore decay heat predictions, adjust operational parameters, and visualize results in real-time. 6) Applications: The framework supports various applications, including safety system validation and spent fuel container optimization. It enables real-time evaluation of worst-case decay heat scenarios, informing the design of passive safety systems and optimizing container designs for long-term storage. Overall, this code provides a robust, accurate, and user-friendly tool for predicting decay heat in MSRs, enhancing reactor safety, and optimizing spent fuel management.

Retamales, Mauricio Eduardo Tano [Idaho National L↗
DOE OSTI2022

Climate Dynamics Preceding Summer Forest Fires in California and the Extreme Case of 2018

Recent record-breaking wildfire seasons in California prompt an investigation into the climate patterns that typically precede anomalous summer burned forest area. Using burned-area data from the U.S. Forest Service’s Monitoring Trends in Burn Severity (MTBS) product and climate data from the fifth major global reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ERA5) over 1984–2018, relationships between the interannual variability of antecedent climate anomalies and July California burned area are spatially and temporally characterized. Lag correlations show that antecedent high vapor pressure deficit (VPD), high temperatures, frequent extreme high temperature days, low precipitation, high subsidence, high geopotential height, low soil moisture, and low snowpack and snowmelt anomalies all correlate significantly with July California burned area as far back as the January before the fire season. Seasonal regression maps indicate that a global midlatitude atmospheric wave train in late winter is associated with anomalous July California burned area. July 2018, a year with especially high burned area, was to some extent consistent with the general patterns revealed by the regressions: low winter precipitation and high spring VPD preceded the extreme burned area. However, geopotential height anomaly patterns were distinct from those in the regressions. Extreme July heat likely contributed to the extent of the fires ignited that month, even though extreme July temperatures do not historically significantly correlate with July burned area. While the 2018 antecedent climate conditions were typical of a high-burned-area year, they were not extreme, demonstrating the likely limits of statistical prediction of extreme fire seasons and the need for individual case studies of extreme years.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Chemometrics and Experimental Design for the Quantification of Nitrate Salts in Nitric Acid: Near-Infrared Spectroscopy Absorption Analysis

Implementing remote, real-time spectroscopic monitoring of radiochemical processing streams in hot cell environments requires efficiency and simplicity. The success of optical spectroscopy for the quantification of species in chemical systems highly depends on representative training sets and suitable validation sets. Selecting a training set (i.e., calibration standards) to build multivariate regression models is both time- and resource-consuming using standard one-factor-at-a-time approaches. This study describes the use of experimental design to generate spectral training sets and a validation set for the quantification of sodium nitrate (0–1 M) and nitric acid (0.1–10 M) using the near-infrared water band centered at 1440 nm. Partial least squares regression models were built from training sets generated by both D- and I-optimal experimental designs and a one-factor-at-a-time approach. The prediction performance of each model was evaluated by comparing the bias and standard error of prediction for statistical significance. D- and I-optimal designs reduced the number of samples required to build regression models compared with one-factor-at-a-time while also improving performance. Models must be confirmed against a validation sample set when minimizing the number of samples in the training set. The D-optimal design performed the best when considering both performance and efficiency by improving predictive capability and reducing number of samples in the training set by 64% compared with the one-factor-at-a-time approach. The experimental design approach objectively selects calibration and validation spectral data sets based on statistical criterion to optimize performance and minimize resources.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗
DOE OSTI2021

Empirical Estimation of Shortest Route Length along U.S. Interstate Highways Based on Great Circle Distance

In this study, 98 regression models were specified for easily estimating shortest distances based on great circle distances along the U.S. interstate highways nationwide and for each of the continental 48 states. This allows transportation professionals to quickly generate distance, or even distance matrix, without expending significant efforts on complicated shortest path calculations. For simple usage by all professionals, all models are present in the simple linear regression form. Only one explanatory variable, the great circle distance, is considered to calculate the route distance. For each geographic scope (i.e., the national or one of the states), two different models were considered, with and without the intercept. Based on the adjusted R-squared, it was observed that models without intercepts generally have better fitness. Additionally, all these models generally have good fitness with the linear regression relationship between the great circle distance and route distance. At the state level, significant variations in the slope coefficients between the state-level models were also observed. Furthermore, a preliminary analysis of the effect of highway density on this variation was conducted.

33 ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2024

Simultaneously improving accuracy and computational cost under parametric constraints in materials property prediction tasks

Abstract Modern data mining techniques using machine learning (ML) and deep learning (DL) algorithms have been shown to excel in the regression-based task of materials property prediction using various materials representations. In an attempt to improve the predictive performance of the deep neural network model, researchers have tried to add more layers as well as develop new architectural components to create sophisticated and deep neural network models that can aid in the training process and improve the predictive ability of the final model. However, usually, these modifications require a lot of computational resources, thereby further increasing the already large model training time, which is often not feasible, thereby limiting usage for most researchers. In this paper, we study and propose a deep neural network framework for regression-based problems comprising of fully connected layers that can work with any numerical vector-based materials representations as model input. We present a novel deep regression neural network, iBRNet, with branched skip connections and multiple schedulers, which can reduce the number of parameters used to construct the model, improve the accuracy, and decrease the training time of the predictive model. We perform the model training using composition-based numerical vectors representing the elemental fractions of the respective materials and compare their performance against other traditional ML and several known DL architectures. Using multiple datasets with varying data sizes for training and testing, We show that the proposed iBRNet models outperform the state-of-the-art ML and DL models for all data sizes. We also show that the branched structure and usage of multiple schedulers lead to fewer parameters and faster model training time with better convergence than other neural networks. Scientific contribution: The combination of multiple callback functions in deep neural networks minimizes training time and maximizes accuracy in a controlled computational environment with parametric constraints for the task of materials property prediction.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2019

Application of a convolutional neural network in permeability prediction: A case study in the Jacksonburg-Stringtown oil field, West Virginia, USA

Permeability is a critical parameter for understanding subsurface fluid flow behavior, managing reservoirs, enhancing hydrocarbon recovery, and sequestering carbon dioxide. In general, permeability is measured in the laboratory based on subsurface core samples, calculated from well logs or estimated from well tests. However, laboratory measurements and well tests are expensive, time-consuming, and usually limited to a few core samples or wells in a hydrocarbon field or carbon storage site. Machine-learning techniques are good options for generating a rapid, robust, and cost-effective permeability prediction model because of their strengths to recognize the potential interrelationships between input and output variables. Convolutional neural networks (CNN), as a good pattern recognition algorithm, are widely used in image processing, natural language processing, and speech recognition, but are rarely used with regression problems and even less often in reservoir characterization. We have developed a CNN regression model to estimate the permeability in the Jacksonburg-Stringtown oil field, West Virginia, which is a potential carbon storage site and enhanced oil recovery operations field. We also evaluate the concept of the geologic feature image, which is converted from geophysical well logs. Five variables, including two commonly available conventional well logs (the gamma rays [GRs] and bulk density) and three well-log-derived variables (the slopes of the GR and bulk density curves, and shale content), are used to generate a geologic feature image. The CNN treats the geologic feature image as the input and the permeability as the desired output. In addition, the permeability predicted using traditional backpropagation artificial neural networks, which are optimized by genetic algorithms and particle swarm optimization, is compared with the permeability estimated using our CNN. Our results indicate that the CNN regression model provides more accurate permeability predictions than the traditional neural network.

Geochemistry & Geophysics↗
DOE OSTI2022

Deep Neural Network Algorithm for CMC Microstructure Characterization and Variability Quantification

Microstructure characterization and variability quantification are crucial for understanding ceramic matrix composites (CMCs) mechanical behavior and deformation mechanisms across length scales. Traditionally, analyses of the micrographs obtained from microscopy are labor-intensive. However, with the vast improvement in computer vision (CV) and deep learning (DL), an automated algorithm can be designed to extract essential microstructure variability from micrographs which can then be used to construct a statistically representative volume element (SRVE). The DL-based algorithm spans the taxonomy of microstructure analyses, including semantic segmentation of microstructure constituents, secondary phases, matrix/fiber interface, and defects, and quantifying the microstructure variability in terms of probability distributions. In this work, C/SiNC and SiC/SiNC CMCs microstructures are semantically segmented through a deep convolutional neural network, followed by variability quantification through the implementation of a fully connected regression layer, hence forming a deep regression network. The deep regression network operates in a feedforward regime, in which the neuron output signal traverses through the network in a unidirectional manner. The weight tensor associated with each layer is updated through a backpropagation stochastic gradient descent approach. The input gray-scale image obtained through in-house scanning electron microscope and confocal microscope micrographs is augmented through affine transformations to increase the training set size, which is then processed through four strided convolutional layers. This compresses the image resolution by half at each layer while increasing the image depth by applying different filters (image encoding). The class activation maps (CAMs) corresponding to the applied filters highlight the key architectural features and assist with the semantic segmentation of the microstructure.

Hamza, Mohamed H.↗
DOE OSTI2017

Integrated Nondestructive Assay Systems to Estimate Plutonium in Spent Fuel Assemblies

An integrated nondestructive assay (NDA) system combining active (neutron generator) and passive neutron detection and passive gamma (PG) detection is being analyzed in order to estimate the amount of plutonium, verify initial enrichment, burnup, and cooling time, and detect partial defects in a spent fuel assembly (SFA). Active signals are measured using the differential die-away (DDA), delayed neutron (DN), and delayed gamma (DG) techniques. Passive signals are measured using total neutron (TN) counts and both gross and spectral resolved gamma counts. To quantify how a system of several NDA techniques is expected to perform, all of the relevant NDA techniques listed above were simulated as a function of various reactor conditions such as initial enrichment, burnup, cooling time, assembly shuffling pattern, reactor operating conditions (including temperature, pressure, and the presence of burnable poisons) by simulating the NDA response for five sets of light water reactor assemblies. In this study, we compare the performance of several exploratory model-fitting options (including neural networks, adaptive regression with splines, iterative bias reduction smoothing, projection pursuit regression, and regression with quadratic terms and interaction terms) to relate data simulated with measurement and model error effects from various subsets of the NDA techniques to the total Pu mass. Isotope masses for SFAs and expected detector responses (DRs) for several NDA techniques are simulated using MCNP, and the DRs become inputs to the fitting process. Such responses include eight signals from DDA, one from DN, one from TN, and up to seven from PG; the DG signal will be examined separately. Results are summarized using the root-mean-squared estimation error for plutonium mass in held-out subsets of the data for a range of model and measurement error variances. Different simulation assumptions lead to different spent fuel libraries relating DRs to Pu mass. Some results for training with one library and testing with another library are also given.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗
DOE OSTI2016

Modeling magnetic fields from a DC power cable buried beneath San Francisco Bay based on empirical measurements

Here, the Trans Bay Cable (TBC) is a ±200-kilovolt (kV), 400 MW 85-km long High Voltage Direct Current (DC) buried transmission line linking Pittsburg, CA with San Francisco, CA (SF) beneath the San Francisco Estuary. The TBC runs parallel to the migratory route of various marine species, including green sturgeon, Chinook salmon, and steelhead trout. In July and August 2014, an extensive series of magnetic field measurements were taken using a pair of submerged Geometrics magnetometers towed behind a survey vessel in four locations in the San Francisco estuary along profiles that cross the cable’s path; these included the San Francisco-Oakland Bay Bridge (BB), the Richmond-San Rafael Bridge (RSR), the Benicia- Martinez Bridge (Ben) and an area in San Pablo Bay (SP) in which a bridge is not present. In this paper, we apply basic formulas that ideally describe the magnetic field from a DC cable summed vectorially with the background geomagnetic field (in the absence of other sources that would perturb the ambient field) to derive characteristics of the cable that are otherwise not immediately observable. Magnetic field profiles from measurements taken along 170 survey lines were inspected visually for evidence of a distinct pattern representing the presence of the cable. Many profiles were dominated by field distortions unrelated to the cable caused by bridge structures or other submerged objects, and the cable’s contribution to the field was not detectable. BB, with 40 of the survey lines, did not yield usable data for these reasons. The unrelated anomalies could be up to 100 times greater than those from the cable. In total, discernible magnetic field profiles measured from 76 survey lines were regressed against the equations, representing eight days of measurement. The modeled field anomalies due to the cable (the difference between the maximum and minimum field along the survey line at the cable crossing) were virtually identical to the measured values. The modeling yielded a pooled cable depth below the bay floor of 2.06 m (±1.46 std dev), and estimated the angle to the horizontal of the imaginary line connecting the crosssectional center of the cable’s two conductors (0.1143 m apart) as 178.9° ±61.9° (std dev) for Ben, 78.6°±37.0° (std dev) for RSR, and 139.9°±27.4° (std dev) for SP. The mean of the eight daily average currents derived from the regressions was 986 ±185 amperes (A) (std dev), as compared to 722 ±95 A (std dev) provided by Trans Bay Cable LLC. Overall, the regressions based on fundamental principles (Biot Savart law) and the vectorial summation of cable and geomagnetic fields provide estimates of cable characteristics consistent with plausible expectations.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2017

Correction of confidence intervals in excess relative risk models using Monte Carlo dosimetry systems with shared errors

Introduction: Measurement errors are ubiquitous in epidemiological studies, especially for doses arising from environmental and occupational exposures, which are difficult to assess. Correction methods for independent measurement errors, have been comprehensively described elsewhere, including regression calibration, simulation extrapolation (SIMEX), etc. However, when exposure estimates are constructed using complicated physical and biological models, uncertainties in the dosimetry system can become very complex. In such systems, uncertainties in some parameters may affect a large group of study participants simultaneously, and hence are “shared” as opposed to independent dosimetric uncertainties. Representing the structure of the uncertainty in the dosimetry by providing multiple realizations of possible dose has been described for several studies. The goal of this paper is to implement, and explore the properties of, a generalized estimating equation-based approach to shared uncertainties based on principles described in an earlier paper. Methods: The effects of shared and unshared dosimetry error on parameter estimates and confidence intervals in a generalized excess relative risk model are examined. A post-hoc correction method is implemented that constructs corrected confidence intervals based on an underlying variance structure described in an earlier paper. In this correction method it is assumed that the estimated slope parameter in a linear excess relative risk model is distributed as a mixture of normal and lognormal components. A simulation study of the behavior of this correction was developed based on multiple realizations provided by the Monte Carlo dosimetry system used for the plutonium exposures in the Mayak Worker Cohort and a risk model with close similarity to that used in analyses of lung cancer data in that cohort. Results: In the simulation study we found that shared dose errors severely degraded the coverage probabilities of confidence limits for the slope parameter in the linear excess relative risk model relating uncertain exposure to lung cancer risk. Using our correction method to adjust the confidence intervals provided much improved coverage probabilities for this parameter. A smaller amount of degradation in confidence interval coverage for other parameters (e.g. those in the baseline and dose modification terms) was noted, and the proposed correction method also improved coverage for these parameters as well. Discussion: Compared to other recently described methods our proposed correction method relies on relatively familiar software to fit dose response models. The basic modeling philosophy described here first uses existing Poisson regression methods to estimate parameters in the excess relative risk model; in this first stage mean dose (computed from the multiple realizations) is used as the independent variable in the regression. Once the model is fit we then correct (widen) the standard errors for all parameters using the proposed correction method. The simulation studies indicate that the methods described here are an important contribution to the available tools for dealing with shared measurement errors in complex risk analysis.

61 RADIATION PROTECTION AND DOSIMETRY↗