Biased Estimates Research Articles

Supplemental structured surveys and pre-existing detection models improve fine-scale density and population estimation with opportunistic community science data

Density and population estimates aid in conservation and stakeholder communication. While free and broadly available community science data can effectively inform species distribution models, they often lack the information necessary to estimate imperfect detection and area sampled, thus limiting their use in fine-scale density modeling. We used structured distance-sampling surveys to model detection probability and calculate survey-specific detection offsets in community science models. We estimated density and population for 16 songbird species under three frameworks: (1) a fixed framework that assumes perfect detection within a specified survey radius, (2) an independent framework that calculates offsets from an independent source, and (3) a calibration framework that calculates offsets from supplemental surveys. Within the calibration framework, we examined the effects of calibration dataset size and data pooling. Estimates of density and population size were consistently biased low in the fixed framework. The independent and calibration frameworks produced reliable estimates for some species, but biased estimates for others, indicating discrepancies in detection probability between structured and community science surveys. The calibration framework produced reliable population estimates with as few as 10 calibration surveys with positive detections. Data pooling dramatically decreased bias. This study provides conservationists and managers with a cost-effective method of estimating density and population.

Probabilistic model for the gravitational wave signal from merging black holes

Parametrized models that predict the gravitational-wave (GW) signal from merging black holes are used to extract source properties from GW observations. The majority of research in this area has focused on developing methods capable of producing highly accurate, point estimate, predictions for the GW signal. A key element missing from every model used in the analysis of GW data is an estimate for how confident the model is in its prediction. This omission increases the risk of biased parameter estimation of source properties. Current strategies include running analyses with multiple models to measure systematic bias however, this fails to accurately reflect the true uncertainty in the models. In this work we develop a probabilistic extension to the phenomenological modeling workflow for nonspinning black holes and demonstrate that the model not only produces accurate point estimates for the GW signal but can be used to provide well-calibrated local estimates for its uncertainty. Our analysis highlights that there is a lack of numerical relativity (NR) simulations available at multiple resolutions which can be used to estimate their numerical error and implore the NR community to continue to improve their estimates for the error in NR solutions published. Waveform models that are not only accurate in their point-estimate predictions but also in their error estimates are a potential way to mitigate bias in GW parameter estimation of compact binaries due to unconfident waveform model extrapolations. Published by the American Physical Society 2024

Data fusion based on short-term memory Kalman filtering using intermittent-displacement and acceleration signal with a time-varying bias

Accurate displacement information plays a vital role in various applications from structural health monitoring to damage detection in physical systems. Due to practical reasons, displacement sensing is often performed using sensors that provide less accuracy or a low sampling rate. This study proposes an algorithm for displacement signal estimation that fuses high sampling rate acceleration data (which consists of unknown time-varying bias) with low sampling rate displacement data. In contrast to the present literature, the state vector in the proposed state-space formulation consists of short-term memory of the past realizations of displacement and velocity, and bias terms which allow (i) online estimation of the displacements by avoiding multi-rate Kalman filtering, and (ii) estimate time-varying bias in the measured acceleration signal. The state equations comprise the finite difference approximations of the time derivative of velocity and displacement which act as a moving average filter to provide smooth estimates, thus preventing the need for offline smoothing procedures. By inclusion of bias terms in the state vector, the drifts in displacement and velocity caused due to time-varying bias in measured acceleration are easily eliminated. The proposed method was found to provide accurate estimates of displacement, velocity, and time-varying bias in measured acceleration signals when tested on different types of bias, including but not limited to constant, linear, and parabolic variations. Additional tests on three oscillators subjected to three earthquake ground motions showed significantly better state estimates for stiff oscillators and improved estimates for flexible oscillators. Experimental validation on a base-excited two-story frame further affirms the capability of the method to estimate the hidden bias in the measured acceleration and provide an accurate estimate of displacement.

A novel LiDAR-GNSS-INS Two-Phase Tightly Coupled integration scheme for precise navigation

Abstract. Recent advances in precise navigation have extensively utilized the integration of Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS), particularly in the domain of intelligent vehicles. However, the efficacy of such navigation systems is considerably compromised by the reflection and multipath disruptions of non-light-of-sight (NLOS) signals. Light Detection and Ranging (LiDAR)-based odometry, an active perception-based sensor known for its precise 3D measurements, has become increasingly prevalent in augmenting navigation systems. Nonetheless, the assimilation of LiDAR odometry with GNSS/INS systems presents substantial challenges. Addressing these challenges, this study introduces a two-phase sensor fusion (TPSF) approach that synergistically combines GNSS positioning, LiDAR odometry, and IMU pre-integration through a dual-stage sensor fusion process. The initial stage employs an Extended Kalman Filter (EKF) to amalgamate the GNSS solution with IMU Mechanization, facilitating the estimation of IMU biases and system initialization. Subsequently, the second stage integrates scan-to-map LiDAR odometry with IMU mechanization to support continuous LiDAR factor estimation. Factor graph optimization (FGO) is then utilized for the comprehensive fusion of LiDAR factors, IMU pre-integration, and GNSS solutions. The efficacy of the proposed methodology is corroborated through rigorous testing on a demanding trajectory from an urbanized open-source dataset, with the system demonstrating a notable enhancement in performance compared to the state-of-the-art algorithms, achieving a translational Standard Deviation (STD) of 1.269 meters.

Total Least Squares Estimation in Hedonic House Price Models

In real estate valuation using the Hedonic Price Model (HPM) estimated via Ordinary Least Squares (OLS) regression, subjectivity and measurement errors in the independent variables violate the Gauss–Markov theorem assumption of a non-random coefficient matrix, leading to biased parameter estimates and incorrect precision assessments. In this contribution, the Errors-in-Variables model equipped with Total Least Squares (TLS) estimation is proposed to address these issues. It fully considers random errors in both dependent and independent variables. An iterative algorithm is provided, and posterior accuracy estimates are provided to validate its effectiveness. Monte Carlo simulations demonstrate that TLS provides more accurate solutions than OLS, significantly improving the root mean square error by over 70%. Empirical experiments on datasets from Boston and Wuhan further confirm the superior performance of TLS, which consistently yields a higher coefficient of determination and a lower posterior variance factor, which shows its more substantial explanatory power for the data. Moreover, TLS shows comparable or slightly superior performance in terms of prediction accuracy. These results make it a compelling and practical method to enhance the HPM.

A constrained maximum likelihood approach to developing well-calibrated models for predicting binary outcomes.

The added value of candidate predictors for risk modeling is routinely evaluated by comparing the performance of models with or without including candidate predictors. Such comparison is most meaningful when the estimated risk by the two models are both unbiased in the target population. Very often data for candidate predictors are sourced from nonrepresentative convenience samples. Updating the base model using the study data without acknowledging the discrepancy between the underlying distribution of the study data and that in the target population can lead to biased risk estimates and therefore an unfair evaluation of candidate predictors. To address this issue assuming access to a well-calibrated base model, we propose a semiparametric method for model fitting that enforces good calibration. The central idea is to calibrate the fitted model against the base model by enforcing suitable constraints in maximizing the likelihood function. This approach enables unbiased assessment of model improvement offered by candidate predictors without requiring a representative sample from the target population, thus overcoming a significant practical challenge. We study theoretical properties for model parameter estimates, and demonstrate improvement in model calibration via extensive simulation studies. Finally, we apply the proposed method to data extracted from Penn Medicine Biobank to inform the added value of breast density for breast cancer risk assessment in the Caucasian woman population.

Does “National Civilized City” policy mitigate air pollution in China? A spatial Durbin difference-in-differences analysis

“National Civilized City” (NCC) is regarded as China’s highest honorary title and most valuable city brand. To win and maintain the “golden city” title, municipal governments must pay close attention to various key appraisal indicators, mainly environmental ones. In this study we verify whether cities with the title are more likely to mitigate SO2 pollution. We adopt the spatial Durbin difference-in-differences (DID) model and use panel data of 283 Chinese cities from 2003 to 2018 to analyze the local (direct) and spillover effects (indirect) of the NCC policy on SO2 pollution. We find that SO2 pollution in Chinese cities is not randomly distributed in geography, suggesting the existence of spatial spillovers and possible biased estimates. Our study treats the NCC policy as a quasi-experiment and incorporates spatial spillovers of NCC policy into a classical DID model to verify this assumption. Our findings show: (1) The spatial distribution of SO2 pollution represents strong spatial spillovers, with the most highly polluted regions mainly situated in the North China Plain. (2) The Moran’s I test results confirms significant spatial autocorrelation. (3) Results of the spatial Durbin DID models reveal that the civilized cities have indeed significantly mitigated SO2 pollution, indicating that cities with the honorary title are acutely aware of the environment in their bid to maintain the golden city brand. As importantly, we notice that the spatial DID term is also significant and negative, implying that neighboring civilized cities have also mitigated their own SO2 pollution. Due to demonstration and competition effects, neighboring cities that won the title ostensibly motivates local officials to adopt stringent policies and measures for lowering SO2 pollution and protecting the environment in competition for the golden title. The spatial autoregressive coefficient was significant and positive, indicating that SO2 pollution of local cities has been deeply affected by neighbors. A series of robustness check tests also confirms our conclusions. Policy recommendations based on the findings for protecting the environment and promoting sustainable development are proposed.

Text Feature Adversarial Learning for Text Generation With Knowledge Transfer From GPT2.

Text generation is a key component of many natural language tasks. Motivated by the success of generative adversarial networks (GANs) for image generation, many text-specific GANs have been proposed. However, due to the discrete nature of text, these text GANs often use reinforcement learning (RL) or continuous relaxations to calculate gradients during learning, leading to high-variance or biased estimation. Furthermore, the existing text GANs often suffer from mode collapse (i.e., they have limited generative diversity). To tackle these problems, we propose a new text GAN model named text feature GAN (TFGAN), where adversarial learning is performed in a continuous text feature space. In the adversarial game, GPT2 provides the "true" features, while the generator of TFGAN learns from them. TFGAN is trained by maximum likelihood estimation on text space and adversarial learning on text feature space, effectively combining them into a single objective, while alleviating mode collapse. TFGAN achieves appealing performance in text generation tasks, and it can also be used as a flexible framework for learning text representations.

Least-cost decarbonization pathways for electricity generation in Finland: A convex quantile regression approach

This study investigates the least-cost decarbonization pathways in the Finnish electricity generation industry in order to achieve the national carbon neutrality goal by 2035. Various abatement measures, such as downscaling production, capital investment, and increasing labor and intermediate inputs, are considered. The marginal abatement costs (MACs) of greenhouse gas emissions are estimated using the convex quantile regression method and applied to unique register-based firm-level greenhouse gas emission data merged with financial statement data. We adjust the MAC estimates for the sample selection bias caused by zero-emission firms by applying the two-stage Heckman correction. Our empirical findings reveal that the median MAC ranges from 0.1 to 3.5 euros per tonne of CO2 equivalent. The projected economic cost of a 90% reduction in emissions is 62 million euros, while the estimated cost of achieving zero emissions is 83 million euros.

Population-Based Denominators Matter: Bias in U.S. Virgin Islands COVID-19 Vaccination Coverage Under Changing Population Counts

PurposeThe U.S. Virgin Islands (USVI) receives an updated population count once every 10 years and used 2010 decennial census population counts to estimate COVID-19 vaccination coverage during the COVID-19 emergency response. We investigated whether using outdated (2010) or modeled (2020 international database [IDB]) population counts biased vaccination coverage estimates used to inform public health priorities during the 2020–2022 COVID-19 response. MethodsWe estimated percentage of USVI residents with a completed primary COVID-19 vaccination series during December 16, 2020–September 20, 2022. Vaccination coverage was calculated as number of persons who completed the vaccination series divided by 2010 and 2020 decennial census population counts and 2020 IDB intercensal estimate. ResultsCOVID-19 vaccination coverage using the 2020 population count was 12 percentage points higher than coverage using 2010 denominator (2010 denominator: 51%; 2020 denominator: 63%). Vaccination coverage estimated using 2020 IDB was approximately equal with the 2010 decennial census estimate (52%). ConclusionsUsing 2010 and modeled population counts underestimated 2020 USVI COVID-19 vaccination coverage given the 18% population decline during 2010–2020, potentially limiting USVI’s ability to assess vaccination progress. Identifying mechanisms for more reliable population enumeration or improved estimate modeling are essential for accurately guiding USVI public health decision-making.

Label-Aware Distribution Calibration for Long-Tailed Classification.

Real-world data usually present long-tailed distributions. Training on imbalanced data tends to render neural networks perform well on head classes while much worse on tail classes. The severe sparseness of training instances for the tail classes is the main challenge, which results in biased distribution estimation during training. Plenty of efforts have been devoted to ameliorating the challenge, including data resampling and synthesizing new training instances for tail classes. However, no prior research has exploited the transferable knowledge from head classes to tail classes for calibrating the distribution of tail classes. In this article, we suppose that tail classes can be enriched by similar head classes and propose a novel distribution calibration (DC) approach named as label-aware DC (LADC). LADC transfers the statistics from relevant head classes to infer the distribution of tail classes. Sampling from calibrated distribution further facilitates rebalancing the classifier. Experiments on both image and text long-tailed datasets demonstrate that LADC significantly outperforms existing methods. The visualization also shows that LADC provides a more accurate distribution estimation.

In silico modelling of stroke volume, cardiac output and systemic vascular resistance in cardiovascular safety pharmacology studies by telemetry

The principle of proportionality of the systolic area of the central aortic pressure to stroke volume (SV) has been long known. The aim of the present work was to evaluate an in silico solution derived from this principle for modelling SV (iSV model) in cardiovascular safety pharmacology studies by telemetry. Blood pressure was measured in the abdominal aorta in accordance with standard practice. Central aortic pressure was modelled from the abdominal aortic pressure waveform using the N-point moving average (NPMA) method for beat-to-beat estimation of SV. First, the iSV was compared to the SV measured by ultrasonic flowmetry in the ascending aorta (uSV) after various pharmacological challenges in beagle dogs anaesthetised with etomidate/fentanyl. The iSV showed minimal bias (0.2 mL i.e. 2%) and excellent agreement with uSV. Then, previous telemetry studies including reference vasoactive and inotropic compounds were retrospectively reanalysed to model drug effects on stroke volume (iSV), cardiac output (iCO) and systemic vascular resistance (iSVR). Among them, the examples of nicardipine and isoprenaline highlight risks of erroneous or biased estimation of drug effects from the abdominal aortic pressure due to pulse pressure amplification. Furthermore, the examples of verapamil, quinidine and moxifloxacin show that iSV, iCO and iSVR are earlier biomarkers than blood pressure itself for predicting drug effect on blood pressure. This in silico modelling approach included in vivo telemetry safety pharmacology studies can be considered as a New Approach Methodology (NAM) that provides valuable additional information and contribute to improving non-clinical translational research to the clinic.

Analyzing the spatial variation characteristics of grid TEC using long-term GIM data

Global ionosphere maps (GIM), a grid model generated based on observations from global GNSS stations, is often used for ionospheric delay correction and total electron content (TEC) analysis, and is an important data for ionospheric modeling and code bias estimation, as well as related applications. Understanding the spatial variation characteristics of the grid TEC in GIM data can help the better use of GIM data. In this paper, the GIM data of CODE (Center for Orbit Determination in Europe) analysis center for a total of 15 years from 2008 to 2022 are used to analyze the variation of TEC of adjacent grid points and the difference of TEC of four grid points in each grid. The results show that the mean values of the grid TEC variations range from about 0.26–1.86 TECu in the longitude direction and 0.36–2.76 TECu in the latitude direction, and most of the differences in the TECs of neighboring grid points are within ± 2 TECu. In the low solar activity years, the STD of the TEC of the four grid points in most of the grids in the GIM data is less than 1.2 TECu, indicating that the TEC values of the four grid points in most of the grids in their GIMs are relatively close to each other and the differences are small, whereas the differences are larger in the high solar activity years. In other words, the TEC values of the areas covered by the GIM grids in the low solar activity years are close. The results of the study can provide some references and lessons for the interpolation and application of the TEC of the GIM grid, and the TEC modeling of the single-station area.

One‐Inflation and Zero‐Truncation Count Data Modelling Revisited With a View on Horvitz–Thompson Estimation of Population Size

SummaryEstimating the size of a hard‐to‐count population is a challenging matter. We consider uni‐list approaches in which the count of identifications per unit is the basis of analysis. Unseen units have a zero count and do not occur in the sample leading to a zero‐truncated setting. Because of various mechanisms, one‐inflation is often an occurring phenomena that can lead to seriously biased estimates of population size. The current work reviews some recent advances on one‐inflation and zero‐truncation modelling, and furthermore focuses here on the impact it has on population size estimation. The zero‐truncated one‐inflated and the one‐inflated zero‐truncated model is compared (also with the model ignoring one‐inflation) in terms of Horvitz–Thompson estimation of population size. The simulation work shows clearly the biasing effect of ignoring one‐inflation. Both models, the zero‐truncated one‐inflated and the one‐inflated zero‐truncated one, are suitable to model ongoing one‐inflation. It is also important to choose an appropriate base‐line distributional model. Finally, all models derived in the paper are illustrated on a number of case studies.

Who is missed in a community-based survey: Assessment and implications of biases due to incomplete sampling frame in a community-based serosurvey, Choma and Ndola Districts, Zambia, 2022.

Community-based serological studies are increasingly relied upon to measure disease burden, identify population immunity gaps, and guide control and elimination strategies; however, there is little understanding of the potential for and impact of sampling biases on outcomes of interest. As part of efforts to quantify measles immunity gaps in Zambia, a community-based serological survey using stratified multi-stage cluster sampling approach was conducted in Ndola and Choma districts in May-June 2022, enrolling 1245 individuals. We carried out a follow-up study among individuals missed from the sampling frame of the serosurvey in July-August 2022, enrolling 672 individuals. We assessed the potential for and impact of biases in the community-based serosurvey by i) estimating differences in characteristics of households and individuals included and excluded (77% vs 23% of households) from the sampling frame of the serosurvey and ii) evaluating the magnitude these differences make on healthcare-seeking behavior, vaccination coverage, and measles seroprevalence. We found that missed households were 20% smaller and 25% less likely to have children. Missed individuals resided in less wealthy households, had different distributions of sex and occupation, and were more likely to seek care at health facilities. Despite these differences, simulating a survey in which missed households were included in the sampling frame resulted in less than a 5% estimated bias in these outcomes. Although community-based studies are upheld as the gold standard study design in assessing immunity gaps and underlying community health characteristics, these findings underscore the fact that sampling biases can impact the results of even well-conducted community-based surveys. Results from these studies should be interpreted in the context of the study methodology and challenges faced during implementation, which include shortcomings in establishing accurate and up-to-date sampling frames. Failure to account for these shortcomings may result in biased estimates and detrimental effects on decision-making.

A stochastic framework for rainfall intensity–timescale–return period relationships. Part I: Theory and estimation strategies

This work presents a stochastic framework for the construction of rainfall intensity–timescale–return period relationships, which was applied in the recent regionalization of design rainfall curves over the Greek territory, described in a companion paper. The methodology outlined herein builds upon the Koutsoyiannis’ et al. 1998 framework, which has been recently revisited and upgraded, and incorporates two different versions: (a) a theoretically consistent stochastic model applicable for rainfall intensity over any scale of interest and (b) a simplified version valid over small scales, which makes parameter estimation easier. Special focus is given to the presentation of the simplified version, which suffices for most engineering tasks. Parameter estimation approaches are presented in detail including the K-moments framework that allows for reliable high-order moment estimation and handling of bias due to spatiotemporal dependence.

A Multistage Nonlinear Method for Aeroengine Health Parameter Estimation Based on Adjacent Operating Points

Abstract Health parameter estimation is the core of engine gas path analysis (GPA), which is widely adopted for engine safety improvement, as well as for operation and maintenance cost reduction. The major challenge of GPA lies in the contradiction between the high dimensions of parameters under estimation, e.g., health parameters, and the limited measurements obtainable from a small number of sensors. Existent GPA methods for health parameters commonly apply dimension reduction before estimation, leading to information loss and hence inaccurate estimation. To tackle the challenge of limited sensor measurements and to have more system outputs than parameters under estimation, we proposed to augment the output vector of the system model by combining the measurements from multiple adjacent operating points. But the engine model can face the problem of homogenization if using data from adjacent operating points. This can, in turn, lead to a low identifiability of parameters. We analyze the internal mechanism of such large deviation of the parameter estimation results based on linear models and argue for the need of nonlinear method. Hence, we propose a multistage nonlinear parameter estimation method for health parameters, combining biased and unbiased estimation. In our extensive simulations based on 10 output measurements of a JT9D engine, our method can estimate 130% more parameters than the widely used GPA method, while reducing the maximum estimation error of health parameters from 2.2% to 0.1%.

Collider bias correction for multiple covariates in GWAS using robust multivariable Mendelian randomization.

Genome-wide association studies (GWAS) have identified many genetic loci associated with complex traits and diseases in the past 20 years. Multiple heritable covariates may be added into GWAS regression models to estimate direct effects of genetic variants on a focal trait, or to improve the power by accounting for environmental effects and other sources of trait variations. When one or more covariates are causally affected by both genetic variants and hidden confounders, adjusting for them in GWAS will produce biased estimation of SNP effects, known as collider bias. Several approaches have been developed to correct collider bias through estimating the bias by Mendelian randomization (MR). However, these methods work for only one covariate, some of which utilize MR methods with relatively strong assumptions, both of which may not hold in practice. In this paper, we extend the bias-correction approaches in two aspects: first we derive an analytical expression for the collider bias in the presence of multiple covariates, then we propose estimating the bias using a robust multivariable MR (MVMR) method based on constrained maximum likelihood (called MVMR-cML), allowing the presence of invalid instrumental variables (IVs) and correlated pleiotropy. We also established the estimation consistency and asymptotic normality of the new bias-corrected estimator. We conducted simulations to show that all methods mitigated collider bias under various scenarios. In real data analyses, we applied the methods to two GWAS examples, the first a GWAS of waist-hip ratio with adjustment for only one covariate, body-mass index (BMI), and the second a GWAS of BMI adjusting metabolomic principle components as multiple covariates, illustrating the effectiveness of bias correction.

Understanding the Impact of Reviews on Consumer Product Choices Under Negotiated Pricing

We study the impact of reviews on product choices when consumers can negotiate price. Although prior research has examined the impact of reviews in various contexts, none has considered a setting in which the purchase price is negotiated. We postulate that online reviews affect demand through consumer baseline product preference and consumer bargaining power. Leveraging a unique data set of consumer reviews of new car purchases, we developed and estimated a structural model to decompose and quantify the two mechanisms of how reviews affect demand: baseline preference mechanism versus bargaining power mechanism. Our counterfactual analysis suggests that the bargaining power mechanism is more important than the baseline preference mechanism in explaining how historical price and discount frequency information in reviews affect seller profits. Ignoring the impact of reviews on the negotiated price leads to a biased estimation of the effectiveness of reviews. We call into caution a popular selling practice of using discounts in exchange for positive reviews, as it can backfire when future consumers leverage this discount information to negotiate a lower price. This paper was accepted by Kartik Hosanagar, information systems. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2020.02658 .

A Post-Processing Multipath/NLoS Bias Estimation Method Based on DBSCAN.

Positioning based on Global Navigation Satellite Systems (GNSSs) in urban environments always suffers from multipath and Non-Line-of-Sight (NLoS) effects. In such conditions, the GNSS pseudorange measurements can be affected by biases disrupting the GNSS-based applications. Many efforts have been devoted to detecting and mitigating the effects of multipath/NLoS, but the identification and classification of such events are still challenging. This research proposes a method for the post-processing estimation of pseudorange biases resulting from multipath/NLoS effects. Providing estimated pseudorange biases due to multipath/NLoS effects serves two main purposes. Firstly, machine learning-based techniques can leverage accurately estimated pseudorange biases as training data to detect and mitigate multipath/NLoS effects. Secondly, these accurately estimated pseudorange biases can serve as a benchmark for evaluating the effectiveness of the methods proposed to detect multipath/NLoS effects. The estimation is achieved by extracting the multipath/NLoS biases from pseudoranges using a clustering algorithm named Density-Based Spatial Clustering of Applications with Noise (DBSCAN). The performance is demonstrated using two real-world data collections in multipath/NLoS scenarios for both static and dynamic conditions. Since there is no ground truth for the pseudorange biases due to the multipath/NLoS scenarios, the proposed method is validated based on the positioning performance. Positioning solutions are computed by subtracting the estimated biases from the raw pseudoranges and comparing them to the ground truth.