Use Of Quantile Regression Research Articles

Estimating reference intervals from an IPD meta-analysis using quantile regression

BackgroundReference intervals, which define an interval in which a specific proportion of measurements from a healthy population are expected to fall, are commonly used in medical practice. Synthesizing information from multiple studies through meta-analysis can provide a more precise and representative reference interval than one derived from a single study. However, the current approaches for estimating the reference interval from a meta-analysis mainly rely on aggregate data and require parametric distributional assumptions that cannot always be checked.MethodsWith the availability of individual participant data (IPD), non-parametric methods can be used to estimate reference intervals without any distributional assumptions. Furthermore, patient-level covariates can be introduced to estimate personalized reference intervals that may be more applicable to specific patients. This paper introduces quantile regression as a method to estimate the reference interval from an IPD meta-analysis under the fixed effects model.ResultsWe compared several non-parametric bootstrap methods through simulation studies to account for within-study correlation. Under fixed effects model, we recommend keeping the studies fixed and only randomly sampling subjects with replacement within each study. ConclusionWe proposed to use the quantile regression in the IPD meta-analysis to estimate the reference interval. Based on the simulation results, we identify an optimal bootstrap strategy for estimating the uncertainty of the estimated reference interval. An example of liver stiffness measurements, a clinically important diagnostic test without explicitly established reference range in children, is provided to demonstrate the use of quantile regression in estimating both overall and subject-specific reference intervals.

How neighbourhood effects vary by achievement level

Abstract After long-standing calls for research into ‘for whom’ neighbourhood matters, the literature has recently gained traction, focussing on background characteristics such as gender, race, and socioeconomic background. However, there is still scarce and only indirect evidence on whether neighbourhood effects vary by children’s academic ability. In this paper, we combine the Neighbourhood Choice Model with quantile regressions to investigate whether neighbourhood effects vary by children’s academic proneness. Using Norwegian register data, we demonstrate heterogeneity in the effects of neighbourhood deprivation as a function of students’ proneness to academic achievement. Students with the very lowest and highest academic potential are the least affected by neighbourhood disadvantage, while those with below average—but not at the lowest end—are most strongly affected. Moreover, while girls are less affected than boys, and children from less educated and affluent families are more affected by neighbourhood disadvantage than their more privileged peers, the effects of these observed moderators are considerably smaller than the heterogeneity observed as a function of academic proneness. Overall, our use of quantile regressions opens up new venues for understanding ‘for whom’ neighbourhoods matter, with individual-level academic vulnerability and strengths being an influential source of neighbourhood effects variation.

Intelligent crude oil price probability forecasting: Deep learning models and industry applications

The crude oil price has been subject to periodic fluctuations because of seasonal changes in industrial demand and supply, weather, natural disasters and global political unrest. An accurate forecast of crude oil prices is of utmost importance for decision makers and industry players in the energy sector. Despite this, the volatility of crude oil prices contributes to the uncertainty of the energy industry, which was particularly challenging following the recent global spread of the COVID-19 epidemic and the Russia–Ukraine conflict. This paper proposes a hybrid deep learning (DL) modelling framework to deal with the volatility of crude oil prices, applying ensemble empirical mode decomposition (EEMD), convolutional neural network (CNN) and bidirectional long short-term memory (BiLSTM) integrated with quantile regression (QR); named EEMD-CNN-BiLSTM-QR. Two real-world datasets on crude oil prices from the West Texas Intermediate and Brent Crude Oil markets were employed to validate the EEMD-CNN-BiLSTM-QR hybrid modelling framework. Given that the probability density forecast can capture uncertainty, an in-depth analysis was carried out and prediction accuracy calculated. The findings of this study demonstrate that the proposed EEMD-CNN-BiLSTM-QR DL modelling framework, which uses the probability density forecast method, is superior to other tested models in terms of its ability to forecast crude oil prices. The novelty of this study stems mainly from its use of QR, which allows for the description of the conditional distribution of predicted variables and the extraction of more uncertain information for probability density forecasts.

A robust approach for inference on style analysis coefficients

Style analysis is an asset class factor model aiming at obtaining information on the internal allocation of a financial portfolio and at comparing portfolios with similar investment strategies. The aim of the paper is to investigate the use of quantile regression to draw inferences on style coefficients. In particular, we compare an approximation widely used from practitioners, the Lobosco–Di Bartolomeo solution, with robust estimators based on constrained median regression. The inference process exploits a rolling window procedure based on subsampling theory. Different sets of outliers have been simulated in order to show differences in the efficiency, in the coverage error and in the length of the resulting confidence intervals. The proposed solution shows better performance in presence of outliers in Y, in X, or in X and Y, in terms both of empirical coverage and of interval lengths, i.e. whereas the performance of the classical solution deteriorates.

A framework for recalibrating pedotransfer functions using nonlinear least squares and estimating uncertainty using quantile regression

Pedotransfer functions (PTFs) have been developed for many regions to estimate values missing from soil profile databases. However, globally there are many areas without existing PTFs, and it is not advisable to use PTFs outside their domain of development due to poor performance. Further, developed PTFs often lack accompanying uncertainty estimations. To address these issues, a framework is proposed where existing equation-based PTFs are recalibrated using a nonlinear least squares (NLS) approach and validated on two regions of Canada; this process is coupled with the use of quantile regression (QR) to generate uncertainty estimates. Many PTFs have been developed to predict soil bulk density, so this variable is used as a case study to evaluate the outcome of recalibration. New coefficients are generated for existing soil bulk density PTFs, and the performance of these PTFs is validated using three case study datasets, one from the Ottawa region of Ontario and two from the province of British Columbia, Canada. The improvement of the performance of the recalibrated PTFs is evaluated using root mean square error (RMSE) and the concordance correlation coefficient (CCC). Uncertainty estimates produced using QR are communicated through the mean prediction interval (MPI) and prediction interval coverage probability (PICP) graphs. This framework produces dataset-specific PTFs with improved accuracy and minimized uncertainty, and the method can be applied to other regional datasets to improve the estimations of existing PTF model forms. The methods are most successful with large datasets and PTFs with fewer variables and minimal transformations; further, PTFs with organic carbon (OC) as one of or the sole input variable resulted in the highest accuracy.

How Much Financial Development Accentuates Income Inequality in Central and Eastern European Countries?

Financial development is often associated with significant economic growth, but studies have shown that a high level of financial development can be the cause of deepening income inequality in many countries. The main objective of the proposed study is to identify to what extent financial development influences income inequality in Central and Eastern European Countries (CEEC). Thus, for the model specification we used as dependent variable the Gini coefficient and as independent variable the financial development index. The sample period for the analysis was from 2004 to 2019, restricted by the lack of data on the Gini coefficient in CEECs. Data on the financial development index were collected from International Monetary Fund, and data on the Gini coefficient were extracted from the World Bank’s Poverty and Inequality Platform. The study unravels several contributions. First of all, the use of quantile regression allowed for the examination of the effects of financial development across the entire distribution of income inequality. Second of all, the use of a comprehensive financial development index offered a more robust and comprehensive measure of financial development compared to single indicators. Taking into account that the Gini coefficient must be close to zero, this result was a positive one with, in essence, financial development reducing income inequality in CEECs. Thirdly, the specific focus on CEECs fills a gap in the literature. Finally, the findings of this study have important policy implications. The obtained results indicate a negative causal relationship between financial development and income inequality, emphasizing the fact that the relationship between these two components cannot be generalized for all regions. These might include measures to promote financial inclusion, improve financial literacy, and enhance the stability and efficiency of financial systems. Supporting financial development in CEECs and similar transition economies can be an effective strategy for tackling income inequality.

Population size in QTL detection using quantile regression in genome-wide association studies

The aim of this study was to evaluate the performance of Quantile Regression (QR) in Genome-Wide Association Studies (GWAS) regarding the ability to detect QTLs (Quantitative Trait Locus) associated with phenotypic traits of interest, considering different population sizes. For this, simulated data was used, with traits of different levels of heritability (0.30 and 0.50), and controlled by 3 and 100 QTLs. Populations of 1,000 to 200 individuals were defined, with a random reduction of 100 individuals for each population. The power of detection of QTLs and the false positive rate were obtained by means of QR considering three different quantiles (0.10, 0.50 and 0.90) and also by means of the General Linear Model (GLM). In general, it was observed that the QR models showed greater power of detection of QTLs in all scenarios evaluated and a relatively low false positive rate in scenarios with a greater number of individuals. The models with the highest detection power of true QTLs at the extreme quantils (0.10 and 0.90) were the ones with the highest detection power of true QTLs. In contrast, the analysis based on the GLM detected few (scenarios with larger population size) or no QTLs in the evaluated scenarios. In the scenarios with low heritability, QR obtained a high detection power. Thus, it was verified that the use of QR in GWAS is effective, allowing the detection of QTLs associated with traits of interest even in scenarios with few genotyped and phenotyped individuals.

Machine Fault Detection Using a Hybrid CNN-LSTM Attention-Based Model.

The predictive maintenance of electrical machines is a critical issue for companies, as it can greatly reduce maintenance costs, increase efficiency, and minimize downtime. In this paper, the issue of predicting electrical machine failures by predicting possible anomalies in the data is addressed through time series analysis. The time series data are from a sensor attached to an electrical machine (motor) measuring vibration variations in three axes: X (axial), Y (radial), and Z (radial X). The dataset is used to train a hybrid convolutional neural network with long short-term memory (CNN-LSTM) architecture. By employing quantile regression at the network output, the proposed approach aims to manage the uncertainties present in the data. The application of the hybrid CNN-LSTM attention-based model, combined with the use of quantile regression to capture uncertainties, yielded superior results compared to traditional reference models. These results can benefit companies by optimizing their maintenance schedules and improving the overall performance of their electric machines.

Environmental Effect Evaluation: A Quantile-Type Path-Modeling Approach

The environment is a key element that affects many aspects of our society, including the economy, education and talents. In this article, the main purpose is to provide statistical models, algorithms and quantitative evidence regarding environmental effect evaluation (EEE). To accomplish this investigation, I first establish a theoretical EEE model and then apply a quantile-type path-modeling algorithm in the developed EEE model at different quantile levels. In the real-data analysis, this article investigates hypotheses regarding this theoretical EEE model and illustrates the statistical performances of quantile-type path-modeling EEE estimators through bootstraps. The results mainly illustrate that the environment has indispensable impacts on the economy, education and science and technology talent directly and has indirect effects on scientific infrastructure and science and technology output. Compared with the existing classical path-modeling algorithm, quantile-type path-modeling EEE estimators make full use of quantile regression and then overcome the classical exploration of only average effects. Both the quantile-type EEE model and quantile-type path-modeling algorithm capture changes in the relations among constructs and between the constructs and observed variables, and this helps to analyze the entire distribution of the outcome variables in this EEE model.

Methodological decisions and their impacts on the perceived relations between school funding and educational achievement

Use of quantile regression, analysis at different levels of aggregation, and inclusion of covariates were used to assess how different methodologies produced varying results with contradictory implications for policy interventions regarding the relation between school funding and achievement. Results indicated that significantly different relations existed at various quantiles of the distribution of achievement, both for student and school-level outcomes. Further, significant differences were found when outcomes at the school-level were compared to outcomes for individual students. Finally, the inclusion of SES drastically altered the results of every analysis in this study, indicating the importance of controlling for confounds. Taking all of these findings into account, we find that higher levels of funding at the student-level were associated with smaller gaps in growth in achievement by SES for lower achieving students. The opposite finding was found at the school level, with higher levels of funding at the school level being associated with larger gaps in growth in achievement based on SES. We compare our findings to results of existing studies, framing our results and interpretations thereof in weighted-student funding literature. Discussion on how these findings can be translated into quasi-experimental research designs are also included.

Spatiotemporal Analysis of Extreme Temperature Change on the Tibetan Plateau Based On Quantile Regression

AbstractTemperature extremes have been extensively observed on the Tibetan Plateau (TP), of which the dynamics were mostly monitored by traditional trend analysis based on linear regression. This method however, is often impacted by a quasi‐periodic heterogeneity disturbance in variance, leading to results susceptible to extreme outliers. In this study, we conducted the comprehensive detection of extreme temperature changes using quantile regression on the TP from 100 weather stations from 1979 to 2019. The use of quantile regression enabled us to characterize the entire density function of the extreme temperature time‐series (the daily minimum, mean and maximum temperature) at each location. Our results revealed that the daily minimum temperature was generally warming faster than the daily maximum and mean temperatures with the respective trend interval of [−0.45, 1.29], [−0.1, 0.98], [−0.01, 1.01] °C/decade among the conditional quantile levels. Furthermore, the warming rate of extremely cold (0.05 quantile level) and hot days (0.95 quantile level) was higher than that of the mean condition (0.5 quantile level) in each extreme temperature time‐series. We also found that there was a significant elevation‐dependent warming only in extremely cold days. In addition, results from the teleconnection analysis showed that the Arctic Oscillation had a strong influence on the extreme cold and hot extremes over the northern and southwestern TP region, while the El Niño‐Southern Oscillation mainly modulated temperatures over the southeastern TP area. This study can potentially offer improved understanding of the temperature extremes over TP in the face of global warming.

Ambulance response time in a Brazilian emergency medical service

Research efforts on ambulance response times for Emergency Medical Services (EMS) calls have been made for decades, especially in developed countries, using different techniques and with varying objectives. In Brazil, a developing country, the scarce scientific production on this vital indicator prioritizes scenarios for EMS in cities with more than one million inhabitants. This shows the importance of extending research to the reality of small and medium-sized cities. This paper presents SAMU, the Brazilian EMS that follows the Franco-German emergency medicine model, compiling numbers related to service at the national level. The use of quantile regression allows the identification of the RT for the EMS and helps to explain the effects of factors at the system level, at the patient level, and specific factors on response time intervals of Southwest Paraná SAMU. This specific EMS, characterized as an inter-municipal consortium of prehospital services, is responsible for prehospital emergency care for an approximate population of 635,000 inhabitants in 42 small towns in the State of Paraná in southern Brazil. From the analysis of the records of 12,050 ambulance dispatches, it was possible to identify the average ambulance response time of 14 min and 25 s. The regression model was able to explain the influence of the independent variables at the system level (presumed severity of the emergency, ambulance dispatch time, and ambulance travel time), at the patient level (age, gender, and characteristic of the emergency) and specific variables of the emergency (day of the week and time of day) on the dependent variable response time over the quantiles, showing that the dispatch time, travel time, time of day, service to male patients and critical cases influence the ambulance response time. This work contributes to deepening the understanding of the management of EMS operations in a developing country, allows the comparison of the RT identified in relation to other countries, and identifies factors that impact the RT for other actors directly or indirectly involved. The practical implications are also presented, as well as how the study impacts the decision-making and management process of the EMS in the short, medium and long term.

Use of Quantile Regression with Fukui–Okubo Model for Prediction and Early Warning of Reservoir Bank Slope Failure

Use of Quantile Regression with Fukui–Okubo Model for Prediction and Early Warning of Reservoir Bank Slope Failure

A quantile regression perspective on external preference mapping

External preference mapping is widely used in marketing and R&D divisions to understand the consumer behaviour. The most common preference map is obtained through a two-step procedure that combines principal component analysis and least squares regression. The standard approach exploits classical regression and therefore focuses on the conditional mean. This paper proposes the use of quantile regression to enrich the preference map looking at the whole distribution of the consumer preference. The enriched maps highlight possible different consumer behaviour with respect to the less or most preferred products. This is pursued by exploring the variability of liking along the principal components as well as focusing on the direction of preference. The use of different aesthetics (colours, shapes, size, arrows) equips standard preference map with additional information and does not force the user to change the standard tool she/he is used to. The proposed methodology is shown in action on a case study pertaining yogurt preferences.

Explaining household electricity consumption using quantile regression, decision tree and artificial neural network

Electricity as an energy carrier par excellence has a vital role in economic development. However, even with the transformation of power systems that follows technological development, mastering the electricity consumption dynamics remains a challenge task. Therefore, in order to be able to develop effective energy policies based on accurate projections, the search for better consumption modelling options is nowadays the focus of many researchers. Given the emerging use of quantile regression, decision tree and artificial neural network models for prediction, this study tries to address the issue of their performances in assessing electricity consumption drivers. A sample of data from a household's electricity consumption survey in Cameroon was used as the empirical context for application of these three models through a comparative analysis. Factors related to equipment and their use, household income level, housing structure, residential living, energy-saving behaviour and weather conditions showed a significant influence on electricity consumption. Artificial neural network model proved to be more efficient than quantile regression and decision tree, with root mean square error (RMSE), mean absolute error (MAE), and mean absolute error percentage (MAPE) lower values and a higher determination coefficient (R2) value. It is expected that these results serve as a reference for making a decision when selecting the most accurate approach for better understanding of those drivers that have the greatest impact on energy demand.

Normative blood pressure response to exercise stress testing in children and adolescents

ObjectiveTo describe normative values for blood pressure (BP) response to maximal exercise in children/adolescents undergoing a treadmill stress test.MethodsFrom a retrospective analysis of medical records, patients who had undergone a...

Deep Quantile Regression for Unsupervised Anomaly Detection in Time-Series

Time-series anomaly detection receives increasing research interest given the growing number of data-rich application domains. Recent additions to anomaly detection methods in research literature include deep neural networks (DNNs: e.g., RNN, CNN, and Autoencoder). The nature and performance of these algorithms in sequence analysis enable them to learn hierarchical discriminative features and time-series temporal nature. However, their performance is affected by usually assuming a Gaussian distribution on the prediction error, which is either ranked, or threshold to label data instances as anomalous or not. An exact parametric distribution is often not directly relevant in many applications though. This will potentially produce faulty decisions from false anomaly predictions due to high variations in data interpretation. The expectations are to produce outputs characterized by a level of confidence. Thus, implementations need the Prediction Interval (PI) that quantify the level of uncertainty associated with the DNN point forecasts, which helps in making better-informed decision and mitigates against false anomaly alerts. An effort has been made in reducing false anomaly alerts through the use of quantile regression for identification of anomalies, but it is limited to the use of quantile interval to identify uncertainties in the data. In this paper, an improve time-series anomaly detection method called deep quantile regression anomaly detection (DQR-AD) is proposed. The proposed method go further to used quantile interval (QI) as anomaly score and compare it with threshold to identify anomalous points in time-series data. The tests run of the proposed method on publicly available anomaly benchmark datasets demonstrate its effective performance over other methods that assumed Gaussian distribution on the prediction or reconstruction cost for detection of anomalies. This shows that our method is potentially less sensitive to data distribution than existing approaches.

Exploring the linkage between dew point temperature and precipitation extremes: A multi-time-scale analysis on a semi-arid Mediterranean region

Understanding warming climate implications on precipitation is of crucial importance, especially for areas particularly subjected to climate changes and land use/cover modifications, which could be extremely vulnerable to phenomena typically caused by rainfall extremes, such as floods and landslides.Past decade has been witnessing an increasing interest on simple modeling approaches based on the observation of commonly available meteorological variables and their physical linkages. In particular, based on the well-known thermodynamic Clausius-Clapeyron (CC) equation, it was widely investigated the scaling relation between rainfall extremes and variables representative of the near surface humidity, typically the surface air temperature, through the use of general regression models. In some cases, conventional approaches have shown some evident limitations related to the use of surface temperature as covariate, limited size of the analyzed datasets and some climatic peculiarities of the investigated areas, especially for tropical, arid and semi-arid environments.The use of quantile regression instead of general regression and dew point temperature instead of surface temperatures have recently revealed promising potentialities for overcoming some of the above limitations for wet areas, while it has been scarcely tested in arid and semi-arid regions. The purpose of this study is to analyze the suitability of a quantile regression-based approach in a semi-arid Mediterranean region and to explore the impact of different modeling choices on the estimation of the scaling rate. More specifically, the sensitivity of extreme precipitation to dew point temperature is investigated through a multi-time-scale analysis, performed on a wide regional dataset of Sicily (Italy) including high temporal resolution climatic data from 86 gauges. The role of the considered rainfall accumulation period, conditional quantile and time-lag between precipitation-dew point temperature paired data is investigated considering different spatial and temporal data aggregation.The results reveal scaling rate values always below the theoretical CC-rate. Hourly and sub-hourly rainfall extremes are more sensitive to changes in dew point temperature than longer precipitation. The analysis of the driest and hottest season shows a dew point temperature dependence of extreme rainfall more complex than for the other seasons. Compared to the use of general regression, the scaling relationship observed with the quantile regression approach, here used, is more regular across different gauges and sub-regions, rainfall accumulation periods and seasons, with similar scaling rates, confirming promising potentialities areas also for semi-arid regions.

Factors affecting COVID-19 mortality: an exploratory study

PurposeThe purpose of this paper is to study the factors affecting COVID-19 mortality.Design/methodology/approachAn empirical model is developed in which the mortality rate per million is the dependent variable, and life expectancy at birth, physician density, education, obesity, proportion of population over the age of 65, urbanization (population density) and per capita income are explanatory variables. Crosscountry data from 184 countries are used to estimate the quantile regression that is employed.FindingsThe estimated results suggest that obesity, the proportion of the population over the age of 65 and urbanization have a positive and statistically significant effect on COVID-19 mortality. Not surprisingly, per capita income has a negative and statistically significant effect on COVID-19 death rate.Research limitations/implicationsThe study is based on the COVID-19 mortality data from June 2020, which have constantly being changed. What data reveal today may be different after two or three months. Despite this limitation, it is expected that this study will serve as the basis for future research in this area.Practical implicationsSince the findings suggest that obesity, population over the age of 65 and density are the primary factors affecting COVID-19 death, the policy-makers should pay particular attention to these factors.Originality/valueTo the authors’ knowledge, this is first attempt to estimate the factors affecting the COVID-19 mortality rate. Its novelty also lies in the use of quantile regressions, which is more efficient in estimating empirical models with heterogeneous data.

The benefits are at the tail: Uncovering the impact of macroprudential policy on growth-at-risk

I uncover heterogeneous effects of macroprudential policy on GDP growth distribution by bringing together the literature on the impact of macroprudential policy and recent developments on the use of quantile regressions. I identify important benefits of macroprudential policy on the left-tail of the GDP growth distribution which contrast with the negative effects found in previous studies using conditional mean models. These benefits may offset the deterioration on growth-at-risk produced by the build-up of cyclical vulnerabilities and the materialization of financial crises. I also find that the impact of macroprudential policy is dependent on the position in the financial cycle, the type of instrument, and the time elapsed since its implementation. In particular, tightening capital measures during expansions may take up to two years to show evidence of benefits on growth-at-risk, while the positive impact of borrower-based measures is rapidly observed. Conversely, in downturns the benefits of loosening capital measures are more immediate, while those of borrower-based measures are limited. This suggests the importance of timing in macroprudential policy. Overall, this study provides a useful framework to assess the impact of macroprudential policy in terms of GDP growth and to identify the term-structure of specific instruments.