Search Results (294)

This study audits whether regime weighting can sharpen conformal forecast intervals without using target-period information or omitting the required finite-sample correction. Conformal prediction builds such intervals from past forecast errors. A natural refinement gives more weight to errors from days whose volatility resembles the forecast day. On daily natural-rubber prices, the refinement appears to work: intervals become about 20% narrower than plain split-conformal, with a significantly better Winkler score. This paper asks whether that gain is real. Three implementation choices are examined, one at a time. The first uses a regime signal that already sees the price move it is meant to predict. The second estimates the regime model on the same residuals the interval is calibrated on. The third omits a correction that the weighted quantile requires in finite samples. The forecast-feasible construction—predictive regime probabilities, a validation-fitted regime model, and the finite-sample correction—shows no detectable improvement over split-conformal, at a paired Winkler difference of $+0.06$ (95% CI $-0.15$ to $+0.17$). Applying the correction alone is not always enough: it removes the apparent advantage on the VMD-augmented ridge residuals, but the filtered comparison arm survives it on the AR(1) residuals at $-0.54$. Only withholding target-period information eliminates the artifact on both. Concentrated weights also leave some intervals unbounded, whereas adaptive conformal baselines remain finite throughout. A controlled simulation reproduces the same apparent gain where no regime information exists at all. Apparent sharpness must therefore be audited for information timing, calibration reuse, and the finite-sample correction. Full article

Annular fillet welds on guide rods lie in stress concentration zones and are susceptible to fatigue failure under dynamic suspension loads, necessitating accurate quality inspection. However, their complex three-dimensional topography makes it difficult for conventional methods to balance detection accuracy and efficiency. This paper presents a geometry-driven 3D point-cloud analysis framework integrating weighted geometric template matching, unified height referencing, and decoupled dual-branch defect discrimination. The framework first locates the weld region via weighted template matching with a dynamic early-termination strategy and establishes a unified height datum through multi-stage filtering and RANSAC plane fitting. Defects are then decoupled by height characteristics: extreme height screening isolates oxide inclusions (ISO 6520-1 No. 303), while quantile dual-threshold layered analysis distinguishes insufficient throat thickness (ISO 6520-1 No. 5213) and excessive convexity (ISO 6520-1 No. 503). Experiments on 100 welds sampled across five production batches, evaluated under EN ISO 5817 quality level C, achieve zero false positives for oxide inclusions and insufficient-throat cases and an 88.2% detection rate for excessive convexity, delivering robust performance under the tested conditions. Full article

Children are vulnerable to indoor environmental exposures, yet evidence from direct measurements in residential environments remains limited. This study included 115 observations from 74 children in Shenzhen during 2023–2024, with 111 observations available for regression-based analyses. Indoor environmental indicators were evaluated using single-indicator generalized estimating equation (GEE) models, restricted cubic spline (RCS) analyses, and weighted quantile sum (WQS) regression. The primary fully adjusted GEE models included age, sex, body mass index, family history of allergy, indoor temperature, district, and survey year. Asthma-related and allergic rhinitis-related symptoms were reported in 35.7% and 47.8% of the 115 observations, respectively. In the fully adjusted models, illuminance was nominally positively associated with both outcomes, whereas NO₂ was nominally inversely associated with asthma-related symptoms. For allergic rhinitis-related symptoms, NH₃ showed a nominal positive association, whereas O₃ and total volatile organic compounds showed nominal inverse associations. However, none of these associations remained statistically significant after false discovery rate correction. Exploratory RCS analyses suggested potential nonlinear associations for several indicators. Exploratory WQS regression did not identify a significant overall mixture association. Overall, the associations were sensitive to adjustment and multiple-testing correction and should be considered exploratory. Longitudinal studies with repeated exposure measurements are needed to confirm these findings. Full article

Financial returns have heavy tails and nonzero skewness. Machine learning risk models typically return isolated quantiles or rest on thin-tailed laws. We introduce a skewed, heavy-tailed distribution that is as easy to use as the normal and that converts any machine learning forecast of conditional moments into a full density. The law splices the left half of one logistic density onto the right half of another. A prescribed mean, variance, and skewness map into its three parameters by elementary algebra. Value at Risk (VaR), Expected Shortfall (ES), optimal holdings, and risk premia then have closed-form expressions. The attainable third-moment interval is wider than that of the smooth half-normal law and even a small departure from symmetry already moves the implied tails away from the Gaussian benchmark. The logistic base has a kurtosis of $4.2$ and above, so tail-risk estimates are more conservative than those of thin-tailed alternatives. Gradient-boosted trees predict the conditional mean, volatility, and skewness that enter the closed-form formulas. The resulting one-day-ahead VaR and ES forecasts are well calibrated and pass standard coverage tests. Unlike quantile-based machine learning forecasts, they deliver the entire conditional density in analytic form. Exponentially weighted moving average moments, fed through the same formulas, already give accurate ES forecasts. An application to stock-index, commodity, and foreign-exchange returns shows that the distribution tracks sample asymmetry and tail behavior. A three-moment calibration matches mean, variance, and skewness. The implied kurtosis is that of the logistic base and is not a free parameter. Full article

Background. Human populations are exposed to complex chemical mixtures, making interaction detection a central challenge in environmental epidemiology. We benchmarked methods for prediction and recovery of interaction structure. Methods. Eight approaches—main-effects Lasso (glmnet_main), interaction Lasso (glmnet_int), hierNet, Random Forests, Bayesian Kernel Machine Regression (BKMR), quantile g-computation (qgcomp), weighted quantile sum regression (gWQS) and SuperLearner—were evaluated across eight linear/nonlinear, additive/interaction, continuous/binary data-generating processes (500 replicates each). Every method completed in all 500 replicates of all eight scenarios. Prediction was assessed on held-out test data using observed-outcome and oracle-referenced metrics; interaction detection was assessed against three known pairwise interactions among 45 candidate pairs, using both hard selection and a threshold-free ranking criterion. BKMR was evaluated at 2000 versus 25,000 MCMC iterations with multi-chain convergence diagnostics. Sensitivity analyses varied sample size, exposure correlation, signal strength, and interaction form. BKMR was also applied illustratively to six metals and prevalent diabetes in NHANES. Results. In additive settings, observed-outcome prediction was similar across methods, but oracle-referenced continuous-outcome error differed by up to six-fold. With interactions, interaction-aware methods clearly outperformed additive-only approaches on the continuous oracle-referenced metrics: in LMI, the oracle MSE was 1.57 for hierNet and 1.87 for glmnet_int against 3.80 for glmnet_main and 4.94 for qgcomp. glmnet_int and hierNet showed comparable sensitivity; hierNet had a modestly lower mean per-replicate false discovery proportion in paired comparisons, while pooled false discovery favored hierNet in the continuous scenarios and glmnet_int in the binary ones; pooled false discovery rates were 0.79 to 0.82 in every interaction scenario, so roughly four in five selected pairs were false. In the scenarios without true interactions, the pooled false discovery rate was exactly 1. Under threshold-free ranking, BKMR was competitive with the penalized methods (pair-ranking AUC: 0.758 to 0.781 across the four interaction scenarios). BKMR’s apparent instability at 2000 iterations reflected inadequate sampling: 93% of monitored parameters had a Gelman–Rubin statistic above 1.1 and the minimum effective sample size was 7.5, whereas at 25,000 iterations the median statistic was 1.02 and the oracle MSE in LMI fell from 6.38 to 2.08. In NHANES, lead, manganese, and iron had the highest posterior inclusion probabilities, with predominantly nonlinear exposure–response functions. Conclusions. Method choice matters most when interactions are present. Interaction-aware methods are preferable when joint effects are relevant, selected interactions require replication given the high false discovery burden, and BKMR comparisons should report sampling budgets and convergence diagnostics rather than treating a short chain as characteristic of the method. Full article

Online conformal prediction methods such as Adaptive Conformal Inference (ACI) and Fully Adaptive Conformal Inference (FACI) adjust prediction intervals under distribution shift, but their calibration is based on a common stream of recent nonconformity scores. We introduce Population-based Adaptive Conformal Ensembles (PACE), a heuristic method that maintains online conformal quantile calibrators with different window sizes, decay rates, and quantile scales. PACE combines the best-calibrated members through fitness-weighted top-K averaging and periodically refreshes the population using clonal selection. For context, Strongly Adaptive Online Conformal Prediction (SAOCP) is a benchmark method that combines online calibration experts operating over different time intervals and provides a formal strongly adaptive regret guarantee. PACE is heuristic and does not provide an analogous regret or coverage guarantee. We evaluate the method on two synthetic datasets and three real-world time series. Against five adaptive conformal baselines, PACE achieves higher empirical coverage during extreme regimes. Compared with SAOCP, it generally obtains higher coverage by producing wider intervals, resulting in less favorable interval scores on most datasets. Full article

In this paper, we propose a new flexible modification of the Lomax distribution within the Bilal-G family generated through the T-X framework, referred to as the Lomax–Bilal distribution. The proposed model offers greater flexibility for modeling-skewed and heavy-tailed phenomena that frequently arise in survival and reliability studies. A comprehensive set of statistical properties is derived, as moments, order statistics, reliability measures, the quantile function, stochastic ordering, and maximum likelihood estimation. Furthermore, several information measures are obtained to characterize the uncertainty structure of the distribution, namely Shannon entropy, Rényi entropy, extropy, cumulative residual extropy, and generalized weighted extropy. Also, the Lorenz, Bonferroni, Zenga curves and Gini index are presented. The practical applicability and effectiveness of the proposed distribution are illustrated through analyses of two real datasets: survival times of patients with head and neck cancer treated with chemotherapy and radiation therapy, and repair times of an airborne communication transceiver. The empirical results showed that the Lomax–Bilal distribution consistently provides a better fit than several well-established lifetime distributions according to goodness-of-fit statistics and information criteria, particularly in modeling tail behavior. These findings suggest that the Lomax–Bilal distribution constitutes a flexible and competitive alternative for analyzing complex lifetime data in reliability engineering and medical survival studies. Full article

Accurate daily gridded precipitation estimates are needed to characterize spatial rainfall variability and support retrospective hydrological analysis in Southeast Asia, where complex rainfall regimes and sparse rain gauge coverage remain major challenges. This study develops an Extended Triple Collocation Analysis (ETCA)-based merged daily precipitation dataset for Southeast Asia using 15 multi-source gridded precipitation products for 2000–2014. The products include gauge-based, satellite-based, reanalysis-based, and merged datasets, all regridded to a common 10 km × 10 km grid. ETCA was applied to all 455 possible three-product combinations to estimate product-level reliability, expressed as the squared correlation coefficient and error variance at each grid cell. The results reveal substantial spatial variability in product reliability. The final ETCA-merged product was generated through pixel-wise reliability-based product selection, quantile-based distributional adjustment using the locally highest-ranked product as an internal reference, and equal-weight averaging of the selected adjusted products. Validation against GSOD daily observations shows that the ETCA-merged product achieves lower RMSD of 2.95 mm day⁻¹, compared with 2.96 mm day⁻¹ for the simple all-product ensemble and 3.01 mm day⁻¹ for the ensemble of the five most regionally reliable products. The corresponding temporal correlations are 0.79, 0.81, and 0.78, respectively. The merged product also improves the representation of high-percentile rainfall, although very intense rainfall remains underestimated. Spatial climatology and annual cycle analyses indicate that the ETCA-merged product preserves the main rainfall patterns and seasonal evolution of Southeast Asia while introducing local adjustments based on product reliability. These findings demonstrate that ETCA provides a useful framework for developing uncertainty-informed precipitation datasets in regions with sparse gauge observations and spatially heterogeneous product performance. Full article

Photovoltaic (PV) power interval forecasting models are commonly trained on data dominated by non-ramp samples, which may weaken uncertainty characterization during rapid power changes. This study proposes a ramp-aware quantile regression Temporal Fusion Transformer (RQR-TFT) that jointly estimates PV power quantiles and the probability of a future ramp event. Ramp labels are constructed from the normalized power change between adjacent sampling instants. A shared Temporal Fusion Transformer (TFT) encoder extracts temporal representations from historical PV power and meteorological variables, and two output branches perform quantile forecasting and ramp-event identification. Ramp-sample-weighted quantile loss and positive-class-weighted classification loss are jointly optimized to increase the influence of minority ramp samples. The proposed method is evaluated for 4 h ahead forecasting using measurements collected from a 50 MW PV power station during 2019–2020. For the nominal 90% prediction interval, RQR-TFT achieves a ramp-sample prediction interval coverage probability (${\mathrm{PICP}}_R$) of 0.864, an overall prediction interval normalized average width (PINAW) of 0.209, and an overall normalized interval score (NIS) of 0.365. The area under the precision–recall curve for ramp-event identification is 0.906. The results demonstrate improved ramp-sample coverage and overall interval quality, although ramp-sample coverage remains below the nominal level. Full article

Objective: To compare non-epidural labor, standard epidural analgesia, and dural puncture epidural analgesia with respect to labor progression, obstetric outcomes, analgesia-related outcomes, and neonatal results. Methods: This single-center retrospective observational cohort study included 168 term singleton parturients: 58 without epidural analgesia, 56 with standard epidural analgesia, and 54 with dural puncture epidural analgesia. The primary outcome was the duration of the second stage of labor. Secondary outcomes included first-stage duration, mode of delivery, blood loss, transfusion, analgesia quality, maternal side effects, umbilical venous pH, biochemical acidosis, Apgar scores, neonatal intensive care unit admission, and neonatal resuscitation. Analyses included group comparisons, adjusted models, quantile regression, Firth penalized logistic regression, propensity-score inverse probability weighting, and a nulliparous sensitivity analysis. Results: Second-stage duration was comparable across groups (median 40.0, 45.0, and 48.5 min in the no-epidural, standard epidural, and dural puncture epidural groups, respectively; p = 0.118). Active pushing time was likewise comparable (8.0, 8.5, and 12.0 min, respectively; p = 0.292), and quantile regression, together with a nulliparous sensitivity analysis—in which the second stage was longest in the no-epidural group—did not indicate epidural-related labor prolongation. Compared with standard epidural analgesia, dural puncture epidural analgesia showed numerically lower pain scores and a numerically lower rescue bolus requirement (30/54 [55.6%] vs. 39/56 [69.6%]; p = 0.183), although neither difference reached statistical significance. Umbilical cord venous pH and biochemical acidosis (cord venous pH < 7.20) did not differ significantly among groups (p = 0.266 and p = 0.335, respectively). The only neonatal endpoint that differed between groups was a recorded fetal acidosis flag, which was more frequent in the epidural groups (0%, 7.1%, and 16.7%; p = 0.004) but was not corroborated by objective cord pH, Apgar scores, neonatal resuscitation, or a significant increase in neonatal intensive care unit admission (5.2%, 10.7%, and 16.7%; p = 0.145). Conclusions: Contemporary low-dose standard epidural and dural puncture epidural analgesia were not associated with clinically meaningful prolongation of labor, active pushing time, or major obstetric intervention. Dural puncture epidural analgesia showed a non-significant trend toward more favorable analgesic performance compared with standard epidural analgesia. Objective neonatal acid–base measures (cord venous pH and biochemical acidosis) did not differ significantly between groups; only a recorded fetal acidosis flag was more frequent in the epidural groups, without correlation to objective cord pH or to adverse clinical neonatal outcomes. These results support the obstetric and neonatal safety of contemporary low-dose epidural techniques; the isolated recorded acidosis signal warrants confirmation in prospective studies with standardized cord blood gas sampling. Full article

Liquefaction and dynamic separation of moisture-sensitive solid bulk cargoes may remain latent for much of a voyage and then manifest as a sustained heel, leaving a comparatively short interval for emergency action. This study develops an exploratory reliability-based analysis of accident chronology using a source-traceable registry of 35 casualties and incidents. Eighteen cases provided post-heel information suitable for the principal emergency time reserve analysis; the observations comprised exact, approximate, reconstructed, interval-censored, and right-censored times. Descriptive statistics calculated from the selected central values and censoring bounds yielded a mean emergency time reserve TR of 200.99 min, a median of 192.20 min, and a range of 67.50–335.10 min. In likelihood-based fitting that retained censoring, the Weibull model achieved the lowest AIC (212.51) and BIC (215.18), with Kolmogorov–Smirnov D = 0.097 (p = 0.989). The fitted lower-tail quantiles were Q10 = 105.40 min and Q25 = 147.99 min, substantially shorter than the descriptive mean. Robustness was examined using nonparametric estimators, Akaike-weighted model averaging, source-confidence weighting, leave-one-out analysis, and alternative interval assumptions. The contribution is a reproducible framework for converting heterogeneous casualty narratives into uncertainty-qualified lower-tail time-reserve evidence and non-prescriptive bridge–team decision support. The framework is not a physical stability model and cannot replace ship-specific GM/GZ calculations, approved loading and stability information, or the master’s judgement. Full article

Construction delays remain a persistent challenge in Australian construction projects, contributing to cost escalation, disrupted work sequences, contractual claims, and reduced confidence in project delivery. Although delay causes have been widely investigated, existing studies often provide broad factor lists and prioritise risks using single-dimension or inconsistent scoring approaches. This limits guidance on which delay risks should receive priority attention when project teams face constrained time, cost, and management resources. This study addresses this limitation by quantifying and prioritising 22 validated delay risk factors in Australian construction projects. Probability of occurrence and schedule impact were evaluated as separate judgement dimensions before being integrated into an overall measure of risk criticality. Data were collected from 48 experienced Australian construction professionals. A dual-dimension Fuzzy Best–Worst Method was applied to derive separate ratio-scale weights for probability of occurrence and schedule impact, with dimension-specific consistency screening used to improve judgement reliability. The resulting weights were integrated using a probability–impact formulation and mapped onto a 5 × 5 Probability–Impact Matrix through quantile-based discretisation. A 10,000-iteration Monte Carlo robustness analysis was subsequently conducted to assess the stability of the resulting rankings under alternative expert-selection and weighting scenarios. The results indicate that the delay risks perceived by the participating professionals as having the highest combined probability and schedule impact are predominantly governance-, approval-, and coordination-related, particularly owner late decisions, change-approval delays, owner requirement changes, cost-estimation deficiencies, design-approval delays, and inadequate planning. The Monte Carlo analysis further indicated that the principal risk rankings remained relatively stable under variations in expert aggregation. Overall, the integrated FBWM–PIM framework provides a structured and practically interpretable approach for eliciting and prioritising expert perceptions of construction delay risk and translating them into an actionable classification tool for allocating limited risk management resources. Full article

The existing research has rarely integrated within-province associations, interprovincial spatial linkages, and multidimensional heterogeneity when examining how digitalization is related to increasing household consumption. Using a balanced panel of 30 provincial-level regions in China for 2011–2023, compiled from national and provincial statistical yearbooks (CSMAR) and the Peking University Digital Financial Inclusion Index, this study constructs a 0–1 digital economy development index with entropy-weighted TOPSIS. Two-way fixed effects estimate the average within-province relationship; global and local Moran’s I and a spatial Durbin model evaluate spatial dependence and decompose direct, indirect, and total effects. Panel quantile regressions and alternative spatial weight matrices serve as robustness checks, whereas instrumental variables and double/debiased machine learning provide supplementary identification evidence. Digital economy development is positively associated with consumption upgrading in the baseline model. Under economic distance weights, the direct, indirect, and total effects are all significantly positive, although their structure differs across consumption categories, urban and rural groups, regions, and temporal stages. The findings support combining digital infrastructure with service capacity, skills, consumer protection, and interprovincial governance while avoiding uniform policy prescriptions across regions. Full article

Deep drilling equipment operates under high-load, strong-vibration, intermittent-impact, and variable environmental conditions, which motivate sensing systems that provide low-noise acquisition, synchronized triaxial measurements, local data integrity, and quantitative measurement-chain characterization. This paper presents a Zynq UltraScale+ MPSoC-based triaxial vibration sensing station for deep drilling equipment applications. The modular station integrates conditioned-voltage triaxial accelerometer interfaces, analog signal conditioning, 24-bit simultaneous analog-to-digital conversion, electrical isolation, local solid-state-drive storage, Ethernet/wireless communication, and GPS-disciplined oven-controlled crystal oscillator (OCXO) timing. The programmable logic performs deterministic acquisition, GPS pulse processing, oscillator calibration, and DMA transfer, while the processing system facilitates storage, network communication, device-state management, and host computer interaction. The sensing electronics are evaluated through zero-input noise, an experiment-specific input-amplitude-to-noise ratio, gain linearity, thermal stability, repeatability, and station-to-station local-PPS timing tests. The characterized electronics achieve a mean equivalent input noise of 0.31 microvolts, a test-derived ratio of 135.08 dB, and a mean station-to-station local-PPS falling-edge difference of 0.34 microseconds. A lightweight post-acquisition interpretation workflow using learnable multichannel weighted fusion, a convolutional autoencoder, a training-distribution-based quantile threshold, and an auxiliary classification branch achieves 0.9705 accuracy and 0.9704 F1-score on a public triaxial bearing dataset under the reported protocol. A crane-based experiment evaluates deployment feasibility and the sensing–analysis workflow using controlled operating events and a removable stationary mass disturbance. The results provide an engineering sensing basis for distributed monitoring studies on deep drilling equipment. Full article

This study evaluates whether an offline system can conservatively rerank future learning activity profiles under temporal distribution shift; it does not test whether an intervention improves actual student learning. An activity profile comprises temporally binned activity type proportions, click intensity, and active bin indicators. A multilayer perceptron (MLP) predicts a base profile from the first 30% of a course, seven local residual candidates are transferred from similar training students, unsupported candidates are excluded, and a ridge outcome model fitted with inverse probability weighting (IPW) estimates simulator-defined utility. The proposed support-constrained bootstrap lower quantile rule (SC-LQ) switches only when the empirical 10th percentile of a candidate’s bootstrap gain distribution is positive; this quantity is a ranking statistic, not a calibrated confidence bound. Experiments used five Open University Learning Analytics Dataset (OULAD) courses and frozen semi-synthetic potential utilities. Under natural decision rules, SC-LQ switched for 47.5% of development students and 52.2% of holdout students, whereas IPW argmax switched for 81.8% and 86.9%, respectively. At exactly matched switching coverage, SC-LQ did not significantly improve the mean simulator value over IPW (paired difference 0.00065, 95% confidence interval (CI) [−0.00030, 0.00161], Holm-adjusted p = 0.135) but reduced overall harm by 0.01533 and utility loss above 0.01 by 0.02494. Candidate-level empirical coverage of the q10 (10th-percentile) statistic was only 0.763 with 200 bootstrap models, confirming that SC-LQ provides empirical risk ranking rather than a safety guarantee. The value–risk pattern persisted across temporal resolutions, candidate set sizes, validation-selected anchors, and several outcome and utility models but failed or weakened in deliberately discontinuous and misspecified settings. These findings support conservative fallback as a simulator-tested risk control principle for active OULAD learners, not as evidence of causal learning improvement. Full article