Biased Estimates Research Articles

Bias and excess variance in election polling: a not-so-hidden Markov model

Abstract With historic misses in the 2016 and 2020 US Presidential elections, interest in measuring polling errors has increased. The most common method for measuring directional errors and non-sampling excess variability during a postmortem for an election is by assessing the difference between the poll result and election result for polls conducted within a few days of the day of the election. Analysing such polling error data is notoriously difficult with typical models being extremely sensitive to the time between the poll and the election. We leverage hidden Markov models traditionally used for election forecasting to flexibly capture time-varying preferences and treat the election result as a peek at the typically hidden Markovian process. Our results are much less sensitive to the choice of time window, avoid conflating shifting preferences with polling error, and are more interpretable despite a highly flexible model. We demonstrate these results with data on polls from the 2004 through 2020 US Presidential elections and 1992 through 2020 US Senate elections, concluding that previously reported estimates of bias in Presidential elections were too extreme by 10%, estimated bias in Senatorial elections was too extreme by 25%, and excess variability estimates were also too large.

Robust Demand Estimation With Customer Choice-Based Models for Sales Transaction Data

We develop a novel statistical method to estimate customer choice among a firm’s portfolio of offerings when the firm cannot directly observe customers who choose not to purchase any product. This censored demand problem is prevalent in many industries such as hotels, airlines, and retail. Although several methods have been proposed to address this problem, they require some level of data aggregation across arrivals and/or choice sets, which results in information loss and potentially biased estimates. Therefore, they have limited applicability in an environment where the prices of a firm’s portfolio of offerings vary over time and sometimes even across different customers. Our proposed method combines several desirable properties, which makes it a better fit for realistic datasets where the available choice sets or attributes of the products in the choice sets change over time. We consider two additional types of information for identification of our model parameters: (1) additional mild assumptions on the customers’ utility function, and (2) external information about a firm’s market share. We then develop a robust estimation procedure that accounts for inaccuracies in either information type and let the data determine the best approach. Through Monte Carlo simulations, we show that our approach provides promising predictions of customer choice behavior when compared with other generally used methods and clearly outperforms those methods in scenarios where the product prices change frequently over time. Utilizing a real hotel transaction dataset provided by Oracle Labs, we further illustrate the improved estimation accuracy of our method compared to benchmark methods. Relative to existing approaches for estimating customer choice-based models, our proposed methodology better suits environments employing dynamic pricing and personalized offering practices, such as hospitality or online retailing.

Invited Commentary: Influence of Incomplete Death Information on Cumulative Risk Estimates.

Censoring at death is the only feasible option if death is not recorded and individuals who died simply no longer contribute visits, such as in the setting of Barberio et al (Am J Epidemiol. 2024) before they acquired access to mortality information. Censoring at death is known to lead to biased estimates of the probability of the event of interest before time $t$. Barberio et al show through simulations that this bias increases with increasing mortality. However, when analyzing claims data it is often important to not exclude individuals with shorter life expectancies: an important strength of observational studies is that they allow estimating treatment effects in more varied populations than typically included in randomized clinical trials. We derive an analytic expression for the bias, and provide two upper bounds for the bias. The bounds inform the usefulness of obtaining mortality information. If the probability of death before the event is known to be small, wider confidence intervals can be created using the first bound on the bias; we provide an algorithm. If the bias is large, obtaining mortality information is important. Barberio et al show that obtaining mortality information can be essential in practice.

Robust Function-on-Function Regression: A Penalized Tau-based Estimation Approach

This study introduces a novel penalized estimation method tailored for function-on-function regression models, combining the robustness of the Tau estimator with penalization techniques to enhance resistance to outliers. Function-on-function regression is essential for modeling intricate relationships between functional predictors and response variables across diverse fields. However, traditional methods often struggle with outliers, leading to biased estimates and diminished predictive performance. Our proposed approach addresses this challenge by integrating robust Tau estimation with penalization, promoting both robustness and parsimony in parameter estimation. Theoretical foundations of the penalized Tau estimator within function-on-function regression are discussed, along with empirical validations through simulation studies and an empirical data analysis. By incorporating penalization, our method not only ensures robust estimation of regression parameters but also promotes model simplicity, offering enhanced interpretability and generalization capabilities in functional data analysis.

Scientific logic and spatio-temporal dependence in analyzing extreme-precipitation frequency: negligible or neglected?

Abstract. Statistics is often misused in hydro-climatology, thus causing research to get stuck on unscientific concepts that hinder scientific advances. In particular, neglecting the scientific rationale of statistical inference results in logical and operational fallacies that prevent the discernment of facts, assumptions, and models, thus leading to systematic misinterpretations of the output of data analysis. This study discusses how epistemological principles are not just philosophical concepts but also have very practical effects. To this aim, we focus on the iterated underestimation and misinterpretation of the role of spatio-temporal dependence in statistical analysis of hydro-climatic processes by analyzing the occurrence process of extreme precipitation (P) derived from 100-year daily time series recorded at 1106 worldwide gauges of the Global Historical Climatology Network. The analysis contrasts a model-based approach that is compliant with the well-devised but often neglected logic of statistical inference and a widespread but theoretically problematic test-based approach relying on statistical hypothesis tests applied to unrepeatable hydro-climatic records. The model-based approach highlights the actual impact of spatio-temporal dependence and a finite sample size on statistical inference, resulting in over-dispersed marginal distributions and biased estimates of dependence properties, such as autocorrelation and power spectrum density. These issues also affect the outcome and interpretation of statistical tests for trend detection. Overall, the model-based approach results in a theoretically coherent modeling framework where stationary stochastic processes incorporating the empirical spatio-temporal correlation and its effects provide a faithful description of the occurrence process of extreme P at various spatio-temporal scales. On the other hand, the test-based approach leads to theoretically unsubstantiated results and interpretations, along with logically contradictory conclusions such as the simultaneous equi-dispersion and over-dispersion of extreme P. Therefore, accounting for the effect of dependence in the analysis of the frequency of extreme P has a huge impact that cannot be ignored, and, more importantly, any data analysis can be scientifically meaningful only if it considers the epistemological principles of statistical inference such as the asymmetry between confirmatory and disconfirmatory empiricism, the inverse-probability problem affecting statistical tests, and the difference between assumptions and models.

On the effect of background seismicity in physics-based earthquake simulations

Physics-based simulations have been extensively employed to generate synthetic earthquake catalogs over the past decade. In this kind of simulation, primary known faults are typically modeled, while smaller and intermediate faults are often neglected. As a result, the location of earthquakes in the simulation is confined to the modeled faults. Furthermore, due to the elimination of off-fault seismicity, a complete match of the frequency-magnitude distribution of the synthetic catalog with observed data is not achieved. This paper presents an approach to include off-fault seismicity (or background seismicity) within the simulation. First, background earthquakes are separated from the primary faults’ earthquakes, and a fault section is assigned to each event. By employing a scaling relationship, the dimensions of these fault sections are determined according to the magnitude of the corresponding event. The fault section’s slip rates are estimated based on the observed frequency-magnitude distribution. Combining the background fault model with the primary fault model results in a comprehensive model that accounts for all stress interactions between fault elements within the simulator in actual locations. As a result, a broader range of frequency-magnitude distribution in the synthetic catalog can be matched to the observations. A case study using one of the available simulators, Virtual Quake, on a sample area is presented. The results demonstrate that eliminating off-fault seismicity could lead to an underestimated assessment of the earthquake probabilities for the whole region and a biased estimation of earthquake rates in individual faults.

Stylized Rendering as a Function of Expectation

We propose a generalization of the rendering equation that captures both the realistic light transport of physically-based rendering (PBR) and a subset of non-photorealistic rendering (NPR) stylizations in a principled manner. The proposed formulation is based on the key observation that both classical transport and certain NPR stylizations can be modeled as a function of expectation. Given this observation, we generalize the recursive integrals of the rendering equation to recursive functions of expectation. As estimating functions of expectation can be challenging, especially recursive ones, we provide a toolkit for unbiased and biased estimation comprising prior work, general strategies, and a novel build-your-own strategy for constructing more complex unbiased estimators from simpler unbiased estimators. We then use this toolkit to construct a complete estimator for the proposed recursive formulation, and implement a sampling algorithm that is both conceptually simple and leverages many of the components of an ordinary path tracer. To demonstrate the practicality of the proposed method we showcase how it captures several existing stylizations like color mapping, cel shading, and cross-hatching, fuses NPR and PBR visuals, and allows us to explore visuals that were previously challenging under existing formulations.

Why the Feldstein–Horioka “puzzle” remains unsolved

AbstractWe argue that the 40‐year‐old Feldstein–Horioka “puzzle” should have never been labeled as such. We discuss two problems with the literature. First, we show that the series of investment and saving rates typically used in empirical exercises to test the Feldstein–Horioka thesis are not appropriate. The correct series to properly test it are not collected. Second, we show that the Feldstein–Horioka regression is not a model in the econometric sense, that is, an equation with a proper error term (a random variable). The reason is that by adding the capital account to their regression, one gets the accounting identity that relates the capital account, domestic investment, and domestic saving. This implies that the estimate of the coefficient of the saving rate in the Feldstein–Horioka regression can be thought of as a biased estimate of the same coefficient in the accounting identity, where it has a value of 1. Because the omitted variable is known, we call it pseudo bias. Given that this (pseudo) bias is known to be negative and less than 1 in absolute terms, it should come as no surprise that the Feldstein–Horioka regression yields a coefficient between 0 and 1.

Adaptive Regularized Tri-Factor Non-Negative Matrix Factorization for Cell Type Deconvolution.

Accurate deconvolution of cell types from bulk gene expression is crucial for understanding cellular compositions and uncovering cell-type specific differential expression and physiological states of diseased tissues. Existing deconvolution methods have limitations, such as requiring complete cellular gene expression signatures or neglecting partial biological information. Moreover, these methods often overlook varying cell-type mRNA amounts, leading to biased proportion estimates. Additionally, they do not effectively utilize valuable reference information from external studies, such as means and ranges of population cell-type proportions. To address these challenges, we introduce an Adaptive Regularized Tri-factor non-negative matrix factorization approach for deconvolution (ARTdeConv). We rigorously establish the numerical convergence of our algorithm. Through benchmark simulations, we demonstrate the superior performance of ARTdeConv compared to state-of-the-art semi-reference-based and reference-free methods. In a real-world application, our method accurately estimates cell proportions, as evidenced by the nearly perfect Pearson's correlation between ARTdeConv estimates and flow cytometry measurements in a dataset from a trivalent influenza vaccine study. Moreover, our analysis of ARTdeConv estimates in COVID-19 patients reveals patterns consistent with important immunological phenomena observed in other studies. The proposed method, ARTdeConv, is implemented as an R package and can be accessed on GitHub for researchers and practitioners.

Decreasing Trend Bumped Probability Distribution: its Properties, Simulations and Application

In this paper, a probability model is proposed which engaged the mathematical combination of Lindley distribution and a trigonometric component known as haversine function〖 δ〗_s. The one parameter model prototype sustains the capacity to forecast multimodal decreasing trend sinusoidal outcomes. By nomenclature, the proposed probability model is called Modified Lindley Trigonometric Distribution (MLTD). Some statistical properties studied include the hazard function, mean residual life function, moments, conditional moments and moment generating function, Bonferroni and Lorenz curve, entropy, asymptotic distribution, order statistics, and parameter estimation; where the hazard function specially features a cyclic or periodic bathtub and inverted bathtub shape in chain format. The numerical behavior of the estimates of the average bias and mean square error were examined under Monte Carlo simulation approach; and an applicative simulation is experimented to underscore the parametric behavior of MLTD in data modeling. A real life flood data is used to illustrate the essence of the development.

Benchmarking and Optimization of Methods for the Detection of Identity-By-Descent in High-Recombining Plasmodium falciparum Genomes.

Genomic surveillance is crucial for identifying at-risk populations for targeted malaria control and elimination. Identity-by-descent (IBD) is increasingly being used in Plasmodium population genomics to estimate genetic relatedness, effective population size (N e ), population structure, and signals of positive selection. Despite its potential, a thorough evaluation of IBD segment detection tools for species with high recombination rates, such as P. falciparum, remains absent. Here, we perform comprehensive benchmarking of IBD callers - probabilistic (hmmIBD, isoRelate), identity-by-state-based (hap-IBD, phased IBD) and others (Refined IBD) - using population genetic simulations tailored for high recombination, and IBD quality metrics at both the IBD segment level and the IBD-based downstream inference level. Our results demonstrate that low marker density per genetic unit, related to high recombination relative to mutation, significantly compromises the accuracy of detected IBD segments. In genomes with high recombination rates resembling P. falciparum, most IBD callers exhibit high false negative rates for shorter IBD segments, which can be partially mitigated through optimization of IBD caller parameters, especially those related to marker density. Notably, IBD detected with optimized parameters allows for more accurate capture of selection signals and population structure; IBD-based N e inference is very sensitive to IBD detection errors, with IBD called from hmmIBD uniquely providing less biased estimates of N e in this context. Validation with empirical data from the MalariaGEN Pf 7 database, representing different transmission settings, corroborates these findings. We conclude that context-specific evaluation and parameter optimization are essential for accurate IBD detection in high-recombining species and recommend hmmIBD for quality-sensitive analysis, such as estimation of N e in these species. Our optimization and high-level benchmarking methods not only improve IBD segment detection in high-recombining genomes but also enhance overall genomic analysis, paving the way for more accurate genomic surveillance and targeted intervention strategies for malaria.

BDS-3 full-frequency precise point positioning: models and performance comparison

With the BDS-3 six-frequency integration, there are more choices to implement the precise point positioning (PPP) technology. To take full advantage of the multi-frequency combination, six typical BDS-3 six-frequency PPP models using uncombined observation (UC), five dual-frequency ionospheric-free (IF) combinations (IF2), four triple-frequency IF combinations (IF3), three four-frequency IF combinations (IF4), two five-frequency IF combinations (IF5), and a single six-frequency IF combination (IF6) were constructed and compared. The results indicate that the positioning accuracies of the six models are similar. The static positioning accuracies reach 4, 3–5, and 12–14 mm in the east, north, and up directions, respectively, while the kinematic positioning accuracies are 24–26, 16–17, and 42–43 mm, respectively. Regarding the convergence times, the UC model is slightly worse than the five IF combined models, except for some cases in the up direction. In the static mode, benefiting from the smallest noise amplification factor, the average convergence times of the IF6 model are the shortest, reaching 12.1, 5.5, and 13.4 min in the three directions, respectively, and are 7%, 15%, and 6% shorter than those of the UC model. In the kinematic mode, with the combination of more signals into a single IF combination, the noise level of the IF combined observables gradually decreases, but the convergence times gradually increase. The kinematic average convergence times of the IF2 model are 15.3, 3.6, and 18.0 min in the three directions, which are 6%, 22%, and 10% and 1%, 16%, and 20% shorter than those of the UC and IF6 models, respectively. In addition, the observation residuals, and the estimates of inter-frequency bias, tropospheric zenith wet delay, and receiver clock offsets from different models were compared and analyzed. The specific selection of the model should be based on actual situations.

Photographic grading to evaluate facial cleanliness and trachoma among children in Amhara region, Ethiopia.

Promotion of facial cleanliness is recommended for the elimination of blinding trachoma, largely because of observational studies that have found an association between various measures of facial uncleanliness and trachoma. However, when a field grader assesses both facial cleanliness and trachoma, associations may be biased. Assessment of photographs of the face and conjunctiva by masked graders may provide a less biased estimate of the relationship between facial cleanliness and trachoma. Face photographs, conjunctival photographs, and conjunctival swabs were obtained on a random sample of 0-9-year-old children from each of 40 communities in Amhara region, Ethiopia. Face photographs were assessed for the presence of seven measures of an unclean face (i.e., wet nasal discharge, dry nasal discharge, wet ocular discharge, dry ocular discharge, food, dust/dirt, and flies) by three independent masked photo-graders. Conjunctival photographs were similarly graded in a masked fashion for signs of clinically active trachoma. Conjunctival swabs were processed for Chlamydia trachomatis DNA. Of 2073 children with complete data, 808 (39%) had evidence of clinically active trachoma, 150 (7%) had evidence of ocular chlamydia infection, and 2524 (91%) had at least one measure of an unclean face. Dry ocular discharge had the strongest association with clinically active trachoma (age- and sex-adjusted prevalence ratio [PR] 1.4, 95% CI 1.2-1.6) and ocular chlamydia infection (PR 1.9, 95%CI 1.3-2.9), although significant associations were observed between each of the measures of facial uncleanliness and trachoma. Masked assessment of face and conjunctival photographs confirmed prior observational studies that have noted associations between various measures of facial uncleanliness and trachoma. The causal relationship between facial uncleanliness and trachoma is unclear since many features used to measure facial cleanliness (e.g., ocular discharge, nasal discharge, and flies) could be consequences of antecedent ocular chlamydia infection. NCT02754583, clinicaltrials.gov.

AGE ESTIMATION USING DIAPHYSEAL LONG BONE LENGTHS IN A NON-ADULT SKELETAL SAMPLE FROM GONUR DEPE, TURKMENISTAN

Introduction. The aim of this study is to provide group-specific regression equations for age estimation of immature human skeletal remains younger than 12 years of age from the diaphyseal length of the six long bones. Further, to compare inverse and classical calibration models for age estimation and the suitability of previously published regression methods for archaeological and present-day populations in relation to the analyzed sample. Materials and methods. The studied sample consists of 128 non-adult individuals from Gonur Depe – the major BMAC site in Turkmenistan (2300–1500 BCE). Regression formulae were obtained for each of the six long bones, separately for entire sample (0–12 years of age) and for two subsamples below and above 2 years of age. For each equation the coefficient of determination (R2) and F-statistics were calculated. Residuals were tested for normality of distribution, autocorrelation, homoscedasticity, and the equality of mean to zero (one-sample t-test). Comparison between inverse and classical calibration was provided using paired samples t-test. To assess the applicability of other regression formulae to the studied sample both the mean residuals (MR) and mean of the absolute value of the residuals (MAR) were calculated, as an estimate of bias and accuracy respectively. Results. All regression models showed a strong statistical significance and high R2 value. The slope coefficients of the regression lines of diaphyseal length upon age are greater for the upper limb bones both in the entire sample and two subsamples separately. The lower limb bones are characterized by lower growth rates. In contrast to the inverse calibration, for the classical model the mean standard errors (MSE) were smaller for the upper limb bones rather than for the lower limb bones. For the lower limb bones the standard error of the estimate (SE) was generally smaller in inverse formulae. Comparison of both models for the femur however shows their equal performance. For the exception of proposed and classical calibration formulae all inverse models for femur diaphyseal length show consistent differences from zero in relation to the studied sample. Discussion. The inverse and classical calibration models as a technique for age estimation using diaphyseal long bone lengths are both equally applicable in the studied sample. It is recommended to use the equations for the lower limb bones in the inverse model and for the upper limb bones in the classical model. The Gonur Depe population is characterized by relatively higher growth rates of the upper limb bones and distal limb segments relative to the proximal ones. Most of the previously published inverse calibration models are not recommended for uncritically use due to the high risk of obtaining biased estimates on samples that are different chronologically and/or territorially.

Incorporating Spatial Autocorrelation into GPP Estimation Using Eigenvector Spatial Filtering

Terrestrial gross primary productivity (GPP) is a critical part of land carbon fluxes. Accurately quantifying GPP in terrestrial ecosystems and understanding its spatiotemporal dynamics are essential for assessing the capability of vegetation to absorb carbon from the atmosphere. Nevertheless, traditional remote sensing estimation models often require complex parameters and data inputs, and they do not account for spatial effects resulting from the distribution of monitoring sites. This can lead to biased parameter estimation and unstable results. To address these challenges, we have raised a spatial autocorrelation light gradient boosting machine model (SA-LGBM) to enhance GPP estimation. SA-LGBM combines reflectance information from remote sensing observations with eigenvector spatial filtering (ESF) methods to create a set of variables that capture continuous spatiotemporal variations in plant functional types and GPP. SA-LGBM demonstrates promising results when compared to existing GPP products. With the inclusion of eigenvectors, we observed an 8.5% increase in R2 and a 20.8% decrease in RMSE. Furthermore, the residuals of the model became more random, reducing the inherent spatial effects within them. In summary, SA-LGBM represents the first attempt to quantify the impact of spatial autocorrelation and addresses the limitations of underestimation present in existing GPP products. Moreover, SA-LGBM exhibits favorable applicability across various vegetation types.

Avoid non‐probability sampling to select population monitoring sites: Comment on McClure and Rolek (2023)

Abstract Population monitoring programmes typically rely on sampling because it is impossible to survey all the sites within the study area. In such a situation, the general recommendation to obtain unbiased estimates of population trends is to select monitoring sites using probability sampling. However, site selection not based on probability sampling, such as selecting sites with the largest abundance of individuals at the beginning of the monitoring programme, is common in practice. Nevertheless, these methods carry the risk of obtaining biased trend estimates. Using simulations, McClure & Rolek (2023) investigated whether three non‐probability sampling site selection methods can yield unbiased trend estimates under some specific conditions. For two of these methods, that is selecting high quality sites and selecting sites known to be occupied, the authors conclude that there is a major risk of obtaining biased trend estimates. For the third method, that is selecting sites with the largest initial abundance, they found conditions in which unbiased estimates can be obtained. They conclude that the general recommendation to use probability sampling should be revised. Here, we show that the authors' results, although perfectly correct, do not invalidate this recommendation. First, we point out that the authors made strong assumptions about the populations' functioning in their simulations, especially that inter‐annual variance in abundance is similar for all sites, which is unlikely in most real populations. We show through simple simulations that even slightly relaxing this assumption invalidates the authors' results. We also point out that for most of the hypotheses made by the authors, it is generally not known at the beginning of a study whether they will be respected. Furthermore, the authors did not provide evidence that selecting sites based on high initial abundance leads to more precise trend estimates than probability sampling methods. Therefore, neither the benefits nor the risks of this method are known. We conclude that until evidence is provided that abundance‐based site selection improves estimate precision and the situations in which it provides unbiased estimates are clearly identified, using probability sampling should remain the rule.

Evaluation of modelled versus observed non-methane volatile organic compounds at European Monitoring and Evaluation Programme sites in Europe

Abstract. Atmospheric volatile organic compounds (VOCs) constitute a wide range of species, acting as precursors to ozone and aerosol formation. Atmospheric chemistry and transport models (CTMs) are crucial to understanding the emissions, distribution, and impacts of VOCs. Given the uncertainties in VOC emissions, lack of evaluation studies, and recent changes in emissions, this work adapts the European Monitoring and Evaluation Programme Meteorological Synthesizing Centre – West (EMEP MSC-W) CTM to evaluate emission inventories in Europe. Here we undertake the first intensive model–measurement comparison of VOCs in 2 decades. The modelled surface concentrations are evaluated both spatially and temporally, using measurements from the regular EMEP monitoring network in 2018 and 2019, as well as a 2022 campaign. To achieve this, we utilised the UK National Atmospheric Emissions Inventory to derive explicit emission profiles for individual species and employed a tracer method to produce pure concentrations that are directly comparable to observations. The degree to which the modelled and measured VOCs agree varies depending on the specific species. The model successfully captures the overall spatial and temporal variations of major alkanes (e.g. ethane, n-butane) and unsaturated species (e.g. ethene, benzene) but less so for propane, i-butane, and ethyne. This discrepancy underscores potential issues in the boundary conditions for the latter species and in their primary emissions from, in particular, the solvent and road transport sectors. Specifically, potential missing propane emissions and issues with its boundary conditions are highlighted by large model underestimations and smaller propane-to-ethane ratios compared to the measurement. Meanwhile, both the model and measurements show strong linear correlations among butane isomers and among pentane isomers, indicating common sources for these pairs of isomers. However, modelled ratios of i-butane to n-butane and i-pentane to n-pentane are approximately one-third of the measured ratios, which is largely driven by significant emissions of n-butane and n-pentane from the solvent sector. This suggests issues with the speciation profile of the solvent sector, underrepresented contributions from transport and fuel evaporation sectors in current inventories, or both. Furthermore, the modelled ethene-to-ethyne and benzene-to-ethyne ratios differ significantly from measured ratios. The different model performance strongly points to shortcomings in the spatial and temporal patterns and magnitudes of ethyne emissions, especially during winter. For OVOCs, the modelled and measured concentrations of methanal and methylglyoxal show a good agreement, despite a moderate underestimation by the model in summer. This discrepancy could be attributed to an underestimation of contributions from biogenic sources or possibly a model overestimation of their photolytic loss in summer. However, the insufficiency of suitable measurements limits the evaluation of other OVOCs. Finally, model simulations employing the CAMS inventory show slightly better agreements with measurements than those using the Centre on Emission Inventories and Projections (CEIP) inventory. This enhancement is likely due to the CAMS inventory's detailed segmentation of the road transport sector, including its associated sub-sector-specific emission profiles. Given this improvement, alongside the previously mentioned concerns about the model's biased estimations of various VOC ratios, future efforts should focus on a more detailed breakdown of dominant emission sectors (e.g. solvents) and the refinement of their speciation profiles to improve model accuracy.

Analytical quaternion-based bias estimation algorithm for fast and accurate stationary gyro-compassing

This work introduces a novel approach to Strapdown Inertial Navigation System (SINS) alignment, distinct from recursive methods like Kalman filtering. The proposed methodology expedites bias error calculations by utilizing quaternion-based analytical relationships, which bypasses the slow convergence behavior associated with recursive algorithms, particularly in azimuth angle error estimation. In addition, the proposed approach demonstrates comparable accuracy to traditional fine alignment methods. Simulations and experiments validate that in contrast to the 10-min time requirement of traditional fine alignment methods (for azimuth angle estimation in stationary conditions), the proposed approach achieves the same accuracy within 20 s. However, limitations exist as the algorithm is applicable only in stationary conditions, and necessitating a high-grade IMU capable of measuring the earth’s rotation rate.

Adjustment for sparse data bias in odds ratios: Significance to appraisal of risk of diabetes due to occupational trichlorfon insecticide exposure

BackgroundBias away from the null in odds ratios (OR), aggravated by low power, is a well-known phenomenon in statistics (sparse data bias). Such bias increases in presence of selection of “significant” results on the basis of null hypothesis testing (effect size magnification, ESM). ObjectivesWe seek to illustrate these issues and adjust for suspected sparse data bias in the context of a reported more than doubling of the odds of new onset type 2 diabetes in presence of occupational trichlorfon insecticide exposure reported in the Agricultural Health Study. MethodsWe performed ESM analysis on the crude ORs extracted from the contingency table in the published report, which is done by simulating selected OR given a posited true OR. Next, we applied easily accessible methods that adjust for sparse data bias to the extracted contingency tables, including data augmentation, bootstrap, Firth's regression, and Bayesian methods with weakly informative priors. ResultsDuring the ESM analysis, we observed that there was a reasonable chance that a “statistically significant” OR of around 2.5–2.6 would be observed for true OR of 1.2. Adjustment for sparse data bias revealed that Bayesian methods outperformed alternative approaches in terms of yielding more precise inference, while not making unjustified distributional assumptions about estimates of OR. The OR in the original paper of about 2.5–2.6 was reduced on average to OR of 1.9 to 2.2, with 95% (Bayesian) credible intervals that included the null. DiscussionIt is reasonable to adjust ORs for sparse data bias when the reported association has societal importance, because policy must be informed by the least biased estimates of the effect. We think that such adjustment would lead to a more appropriate evaluation of the extent of evidence on the contribution of occupational exposure to trichlorfon pesticide to risk of new onset diabetes.

ReCRec: Reasoning the Causes of Implicit Feedback for Debiased Recommendation

Implicit feedback ( e.g ., user clicks) is widely used in building recommender systems (RS). However, the inherent notorious exposure bias significantly affects recommendation performance. Exposure bias refers a phenomenon that implicit feedback is influenced by user exposure, and does not precisely reflect user preference. Current methods for addressing exposure bias primarily reduce confidence in unclicked data, employ exposure models, or leverage propensity scores. Regrettably, these approaches often lead to biased estimations or elevated model variance, yielding sub-optimal results. To overcome these limitations, we propose a new method ReCRec that Re asons the C auses behind the implicit feedback for debiased Rec ommendation. ReCRec identifies three scenarios behind unclicked data — i.e. , unexposed, dislike or a combination of both. A reasoning module is employed to infer the category to which each instance pertains. Consequently, the model is capable of extracting reliable positive and negative signals from unclicked data, thereby facilitating more accurate learning of user preferences. We also conduct thorough theoretical analyses to demonstrate the debiased nature and low variance of ReCRec. Extensive experiments on both semi-synthetic and real-world datasets validate its superiority over state-of-the-art methods.