AVERAGE DENSITY ESTIMATORS: EFFICIENCY AND BOOTSTRAP CONSISTENCY

This paper highlights a tension between semiparametric efficiency and bootstrap consistency in the context of a canonical semiparametric estimation problem, namely the problem of estimating the average density. It is shown that although simple plug-in estimators suffer from bias problems preventing them from achieving semiparametric efficiency under minimal smoothness conditions, the nonparametric bootstrap automatically corrects for this bias and that, as a result, these seemingly inferior estimators achieve bootstrap consistency under minimal smoothness conditions. In contrast, several “debiased” estimators that achieve semiparametric efficiency under minimal smoothness conditions do not achieve bootstrap consistency under those same conditions.

Resampling Plans and the Estimation of Prediction Error

This article was prepared for the Special Issue on Resampling methods for statistical inference of the 2020s. Modern algorithms such as random forests and deep learning are automatic machines for producing prediction rules from training data. Resampling plans have been the key technology for evaluating a rule’s prediction accuracy. After a careful description of the measurement of prediction error the article discusses the advantages and disadvantages of the principal methods: cross-validation, the nonparametric bootstrap, covariance penalties (Mallows’ Cp and the Akaike Information Criterion), and conformal inference. The emphasis is on a broad overview of a large subject, featuring examples, simulations, and a minimum of technical detail.

Flood season segmentation and scheme optimization in the Yellow River

River flood season segmentation is a significant measure for flood prevention. This study aims to carry out theoretical analysis on flood season segmentation methods and put forward a framework for proper flood season segmentation through comparison between different segmentation methods. The studied framework consists of a Fisher optimal partition method for determining the optimum numbers of the sub-seasons, an ensemble approach for segmenting a defined flood season, and a nonparametric bootstrap combined with a fuzzy optimum selection method (NB-FOS) for testing the rationality of the flood season staging schemes. The present research findings show that different methods could result in different staging schemes. It is proved through rational analysis that the staging scheme obtained by probability change point (PCP) is superior to others. The flood season of the downstream reach of the Yellow River can be segmented into three sub-seasons, i.e. early flood season (01 June–20 July), main flood season (21 July–28 September), and late flood season (29 September–08 November). The segmentation results of the flood season should play an active role in flood prevention.

Stationarity in the Prices of Energy Commodities. A Nonparametric Approach

In this paper, we address the classical problem of testing for stationarity in the prices of energy-related commodities. A panel of fourteen time series of monthly prices is analyzed for the 1980–2020 period. Nine of the series are classical nonrenewable, GHG-emissions-intensive resources (coal, crude oil, natural gas), whereas the remaining, low-emission group includes both uranium and four commodities employed in biofuels (rapeseed, palm, and soybean oils, and ethanol). A nonparametric, bootstrap-based stationarity testing framework is employed. The main advantage of this procedure is its asymptotically model-free nature, being less sensitive than parametric tests to the risks of misspecification and detection of spurious unit roots, although it has the potential limitation of typically requiring larger samples than mainstream tools. Results suggest that most of the series analyzed may be trend stationary. The only exception would be crude oil, where different conclusions are obtained depending on whether a seasonal correction is applied or not.

Calibration of surgical tools using multilevel modeling with LINEX loss function: Theory and experiment

Quantifying the tool–tissue interaction forces in surgery can be used in the training process of novice surgeons to help them better handle surgical tools and avoid exerting excessive forces. A significant challenge concerns the development of proper statistical learning techniques to model the relationship between the true force exerted on the tissue and several outputs read from sensors mounted on the surgical tools. We propose a nonparametric bootstrap technique and a Bayesian multilevel modeling methodology to estimate the true forces. We use the linear exponential loss function to asymmetrically penalize the over and underestimation of the applied forces to the tissue. We incorporate the direction of the force as a group factor in our analysis. A weighted approach is used to account for the nonhomogeneity of read voltages from the surgical tool. Our proposed Bayesian multilevel models provide estimates that are more accurate than those under the maximum likelihood and restricted maximum likelihood approaches. Moreover, confidence bounds are much narrower and the biases and root mean squared errors are significantly smaller in our multilevel models with the linear exponential loss function.

Using multivariate endophenotypes to identify psychophysiological mechanisms associated with polygenic scores for substance use, schizophrenia, and education attainment

Background To better characterize brain-based mechanisms of polygenic liability for psychopathology and psychological traits, we extended our previous report (Liu et al. Psychophysiological endophenotypes to characterize mechanisms of known schizophrenia genetic loci. Psychological Medicine, 2017), focused solely on schizophrenia, to test the association between multivariate psychophysiological candidate endophenotypes (including novel measures of θ/δ oscillatory activity) and a range of polygenic scores (PGSs), namely alcohol/cannabis/nicotine use, an updated schizophrenia PGS (containing 52 more genome-wide significant loci than the PGS used in our previous report) and educational attainment. Method A large community-based twin/family sample (N = 4893) was genome-wide genotyped and imputed. PGSs were constructed for alcohol use, regular smoking initiation, lifetime cannabis use, schizophrenia, and educational attainment. Eleven endophenotypes were assessed: visual oddball task event-related electroencephalogram (EEG) measures (target-related parietal P3 amplitude, frontal θ, and parietal δ energy/inter-trial phase clustering), band-limited resting-state EEG power, antisaccade error rate. Principal component analysis exploited covariation among endophenotypes to extract a smaller number of meaningful dimensions/components for statistical analysis. Results Endophenotypes were heritable. PGSs showed expected intercorrelations (e.g. schizophrenia PGS correlated positively with alcohol/nicotine/cannabis PGSs). Schizophrenia PGS was negatively associated with an event-related P3/δ component [β = −0.032, nonparametric bootstrap 95% confidence interval (CI) −0.059 to −0.003]. A prefrontal control component (event-related θ/antisaccade errors) was negatively associated with alcohol (β = −0.034, 95% CI −0.063 to −0.006) and regular smoking PGSs (β = −0.032, 95% CI −0.061 to −0.005) and positively associated with educational attainment PGS (β = 0.031, 95% CI 0.003–0.058). Conclusions Evidence suggests that multivariate endophenotypes of decision-making (P3/δ) and cognitive/attentional control (θ/antisaccade error) relate to alcohol/nicotine, schizophrenia, and educational attainment PGSs and represent promising targets for future research.

Ant Colony System Optimization for Spatiotemporal Modelling of Combined EEG and MEG Data

Electroencephalography/Magnetoencephalography (EEG/MEG) source localization involves the estimation of neural activity inside the brain volume that underlies the EEG/MEG measures observed at the sensor array. In this paper, we consider a Bayesian finite spatial mixture model for source reconstruction and implement Ant Colony System (ACS) optimization coupled with Iterated Conditional Modes (ICM) for computing estimates of the neural source activity. Our approach is evaluated using simulation studies and a real data application in which we implement a nonparametric bootstrap for interval estimation. We demonstrate improved performance of the ACS-ICM algorithm as compared to existing methodology for the same spatiotemporal model.

Population Pharmacokinetics and Dose Optimization of Ganciclovir in Critically Ill Children

Objective: The present study aims to establish a population pharmacokinetic model of ganciclovir and optimize the dosing regimen in critically ill children suffering from cytomegalovirus related disease.Methods: A total of 104 children were included in the study. The population pharmacokinetic model was developed using the Phoenix NLME program. The final model was validated by diagnostic plots, nonparametric bootstrap, visual predictive check, and normalized prediction distribution errors. To further evaluate and optimize the dosing regimens, Monte Carlo simulations were performed. Moreover, the possible association between systemic exposure and hematological toxicity were also monitored in the assessment of adverse events.Results: The ganciclovir pharmacokinetics could be adequately described by a one-compartment model with first-order elimination along with body weight and estimated glomerular filtration rate as significant covariates. As showed in this study, the typical population parameter estimates of apparent volume of distribution and apparent clearance were 11.35 L and 5.23 L/h, respectively. Simulations indicated that the current regimen at a dosage of 10 mg/kg/d would result in subtherapeutic exposure, and elevated doses might be required to reach the target ganciclovir level. No significant association between neutropenia, the most frequent toxicity reported in our study (19.23%), and ganciclovir exposure was observed.Conclusion: A population pharmacokinetic model of intravenous ganciclovir for critically ill children with cytomegalovirus infection was successfully developed. Results showed that underdosing of ganciclovir was relatively common in critically ill pediatric patients, and model-based approaches should be applied in the optimizing of empiric dosing regimens.

Nonparametric inference for interventional effects with multiple mediators

Understanding the pathways whereby an intervention has an effect on an outcome is a common scientific goal. A rich body of literature provides various decompositions of the total intervention effect into pathway-specific effects. Interventional direct and indirect effects provide one such decomposition. Existing estimators of these effects are based on parametric models with confidence interval estimation facilitated via the nonparametric bootstrap. We provide theory that allows for more flexible, possibly machine learning-based, estimation techniques to be considered. In particular, we establish weak convergence results that facilitate the construction of closed-form confidence intervals and hypothesis tests and prove multiple robustness properties of the proposed estimators. Simulations show that inference based on large-sample theory has adequate small-sample performance. Our work thus provides a means of leveraging modern statistical learning techniques in estimation of interventional mediation effects.