Search

The present special issue features a collection of papers presented at the 2017 International Panel Data Conference, hosted by the University of Macedonia in Thessaloniki, Greece. The conference marked the 40th anniversary of the inaugural International Panel Data Conference, which was held in 1977 at INSEE in Paris, under the auspices of the French National Centre for Scientific Research. As a collection, the papers appearing in this special issue of the Journal of Econometrics continue to advance the analysis of panel data, and paint a state-of-the-art picture of the field. (c) 2020 Elsevier B.V. All rights reserved.

Associations between anthropometric factors and breast cancer (BC) risk have varied inconsistently by estrogen and/or progesterone receptor (ER/PR) status. Associations between prediagnostic anthropometric factors and risk of premenopausal and postmenopausal BC overall and ER/PR status subtypes were investigated in a pooled analysis of 20 prospective cohorts, including 36,297 BC cases among 1,061,915 women, using multivariable Cox regression analyses, controlling for reproductive factors, diet and other risk factors. We estimated dose–response relationships and tested for nonlinear associations using restricted cubic splines. Height showed positive, linear associations for premenopausal and postmenopausal BC risk (6–7% RR increase per 5 cm increment), with stronger associations for receptor-positive subtypes. Body mass index (BMI) at cohort baseline was strongly inversely associated with premenopausal BC risk, and strongly positively—and nonlinearly—associated with postmenopausal BC (especially among women who never used hormone replacement therapy). This was primarily observed for receptor-positive subtypes. Early adult BMI (at 18–20 years) showed inverse, linear associations for premenopausal and postmenopausal BC risk (21% and 11% RR decrease per 5 kg/m2, respectively) with stronger associations for receptor-negative subtypes. Adult weight gain since 18–20 years was positively associated with postmenopausal BC risk, stronger for receptor-positive subtypes, and among women who were leaner in early adulthood. Women heavier in early adulthood generally had reduced premenopausal BC risk, independent of later weight gain. Positive associations between height, baseline (adult) BMI, adult weight gain and postmenopausal BC risk were substantially stronger for hormone receptor-positive versus negative subtypes. Premenopausal BC risk was positively associated with height, but inversely with baseline BMI and weight gain (mostly in receptor-positive subtypes). Inverse associations with early adult BMI seemed stronger in receptor-negative subtypes of premenopausal and postmenopausal BC. Electronic supplementary material The online version of this article (10.1007/s10654-020-00688-3) contains supplementary material, which is available to authorized users.

This article studies estimation of linear panel regression models with heterogeneous coefficients using a class of weighted least squares estimators, when both the regressors and the error possibly contain a common latent factor structure. Our theory is robust to the specification of such a factor structure because it does not require any information on the number of factors or estimation of the factor structure itself. Moreover, our theory is efficient, in certain circumstances, because it nests the GLS principle. We first show how our unfeasible weighted-estimator provides a bias-adjusted estimator with the conventional limiting distribution, for situations in which the OLS is affected by a first-order bias. The technical challenge resolved in the article consists of showing how these properties are preserved for the feasible weighted estimator in a double-asymptotics setting. Our theory is illustrated by extensive Monte Carlo experiments and an empirical application that investigates the link between capital accumulation and economic growth in an international setting. Supplementary materials for this article are available online.

In survival analysis, it often happens that a certain fraction of the subjects under study never experience the event of interest, that is, they are considered "cured." In the presence of covariates, a common model for this type of data is the mixture cure model, which assumes that the population consists of two subpopulations, namely the cured and the non-cured ones, and it writes the survival function of the whole population given a set of covariates as a mixture of the survival function of the cured subjects (which equals one), and the survival function of the non-cured ones. In the literature, one usually assumes that the mixing probabilities follow a logistic model. This is, however, a strong modeling assumption, which might not be met in practice. Therefore, in order to have a flexible model which at the same time does not suffer from curse-of-dimensionality problems, we propose in this paper a single-index model for the mixing probabilities. For the survival function of the non-cured subjects we assume a Cox proportional hazards model. We estimate this model using a maximum likelihood approach. We also carry out a simulation study, in which we compare the estimators under the single-index model and under the logistic model for various model settings, and we apply the new model and estimation method on a breast cancer data set. ispartof: BIOMETRICS vol:75 issue:2 pages:452-462 ispartof: location:United States status: published

In socioeconomics or in Biological studies, observations on individuals are often observed longitudinally on a Likert-type scale with substantially large proportion of zeros. This leads to a special case of mixture structured data where extra-variation occurs. Obviously the standard ordinal data analysis fails to provide appropriate statistical inference. We propose a suitable zero inflated semiparametric ordinal model that takes into account the non linear link between the ordinal response and a covariate. A sieve maximum likelihood estimator(MLE) is proposed for the regression parameter of interest. We also propose a test for the zero proportion in this semiparametric model. A simulation study has been carried out to investigate the performance of the estimator as well as the test. We illustrate the methodology using data from a survey on Tuberculosis patients in and around Kolkata, India. (C) 2018 Elsevier B.V. All rights reserved.

A new semiparametric time series model is introduced - the semiparametric transition (SETR) model - that generalizes the threshold and smooth transition models by letting the transition function to be of an unknown form. Estimation is based on a combination of the (local) least squares estimations of the transition function and regression parameters. The asymptotic behavior for the regression coefficient estimator of the SETR model is established, including its oracle property. Monte Carlo simulations demonstrate that the proposed estimator is more robust to the form of the transition function than parametric threshold and smooth transition methods and more precise than varying coefficient estimators.

Deep learning (DL), Artificial Neural Network (ANN), Kernel Nearest Neighbor (k-NN), and Support Vector Machine (SVM) have been applied to numerous fields owing to their high-accuracy and ability to analyze the non-linear problems. In this study, these machine learning algorithms (MLAs) are used to predict emission and performance characteristics of a CI engine fuelled with various metal-oxide based nano particles (Al2O3, CuO, and TiO2) at a mass fractions of 200 ppm. Assessed parameters in the study are carbon dioxide (CO), nitrogen oxide (NOx), exhaust gas temperature (EGT), brake specific fuel consumption (BSFC), and brake thermal efficiency (BTE). To evaluate the success of algorithms, four metrics (R-2, RMSE, rRMSE, and MBE) are discussed in detail. Tests performed at varying engine speeds from 1500 rpm to 3400 rpm with the intervals of 100 rpm. The addition of nanoparticles simultaneously reduced CO and NOx emissions because they ensured more complete combustion thanks to their inherent oxygen, the higher surface to volume ratio, superior thermal conductivities and their catalytic activity role. Further, the nano-sized particles ensured an accelerated heat transfer from the combustion chamber. In comparison with that of neat diesel fuel, the reduction in NOx is found to be 3.28, 7.53, and 10.05%, and the reduction in CO is found to be 8.3, 11.6, and 15.5% for TiO2, Al2O3, and CuO test fuels, respectively. Moreover, the presence of nanoparticles in test fuels has improved engine performance. As compared with those of neat diesel fuel, the doping of nanoparticles drops the BSFC value by 5.54, 7.89, and 9.96% for TiO2, CuO, and Al2O3, respectively, and enhanced BTE value to be 6.15, 8.87, and 11.23% for TiO2, CuO, and Al2O3, respectively. On the other hand, it can be said that all algorithms presented very satisfying results in the prediction of CI engine responses. All R-2 has changed between 0.901 and 0.994, and DL has given the highest R-2 value for each engine response. In terms of rRMSE, all results (except for one result in k-NN) are categorized as excellent according to the classification in the literature. Considering all metrics together, DL is giving the best results in the prediction of engine responses for the dataset used in this paper. Then it is closely followed by ANN, SVM, and k-NN algorithms, respectively. In conclusion, this paper is proving that the nanoparticle addition for ICEs is significantly dropping the exhaust pollutants, and improving the engine performance, and further the results can be successfully predicted with the machine learning algorithms. (c) 2020 Elsevier Ltd. All rights reserved. Duzce University Scientific Research Projects Coordination UnitDuzce University [2020.07.04.1097, 2019.07.04.1049]; Duzce UniversityDuzce University This present work is being funded by Duzce University Scientific Research Projects Coordination Unit with the project numbers 2020.07.04.1097, and 2019.07.04.1049. The authors thank Duzce University for its financial support, and express special thanks to PhD Candidate Melahat Sevgul Bakay from Biomedical Engineering Department at Duzce University for her valuable suggestions in the prediction stages of the paper. WOS:000596172500006 2-s2.0-85094937244

Introduction, Evidence suggests that the built environment can influence the intensity of physical activity. However, despite the importance of the geographic context, most of the studies do not consider the spatial framework of this association. We aimed to assess individual spatial dependence of objectively measured moderate and vigorous physical activity (MVPA) and describe the characteristics of the built environment among spatial clusters of MVPA., Methods, Cross-sectional data from the second follow-up (2014-2017) of CoLaus vertical bar PsyCoLaus, a longitudinal population-based study of the Lausanne area (Switzerland), was used to objectively measure MVPA using accelerometers. Local Moran's I was used to assess the spatial dependence of MVPA and detect geographic clusters of low and high MVPA. Additionally, the characteristics of the built environment observed in the clusters based on raw MVPA and MVPA adjusted for socioeconomic and demographic factors were compared., Results, Data from 1,889 participants (median age 63, 55% women) were used. The geographic distribution of MVPA and the characteristics of the built environment among clusters were similar for raw and adjusted MVPA. In the adjusted model, we found a low concentration of individuals within spatial clusters of high MVPA (median: 38.5mins; 3% of the studied population) and low MVPA (median: 10.9 mins; 2% of the studied population). Yet, clear differences were found in both models between clusters regarding the built environment; high MVPA clusters were located in areas where specific compositions of the built environment favor physical activity., Conclusions, Our results suggest the built environment may influence local spatial patterns of MVPA independently of socioeconomic and demographic factors. Interventions in the built environment should be considered to promote physically active behaviors in urban areas.

Purpose The purpose of this study is to evaluate socioeconomic determinants of fecundability. Methods Among 8654 female pregnancy planners from Pregnancy Study Online, a North American prospective cohort study (2013–2019), we examined associations between socioeconomic status and fecundability (the per-cycle probability of conception). Information on income and education was collected via baseline questionnaires. Bimonthly follow-up questionnaires were used to ascertain pregnancy status. We estimated fecundability ratios (FRs) and 95% confidence intervals (CIs) using proportional probabilities regression, controlling for potential confounders. Results Relative to an annual household income of greater than or equal to $150,000, adjusted FRs were 0.91 (95% CI: 0.83–1.01) for less than $50,000, 0.99 (95% CI: 0.92–1.07) for $50,000–$99,000, and 1.09 (95% CI: 1.01–1.18) for $100,000–$149,000. FRs for less than 12, 13–15, and 16 years of education, relative to greater than or equal to 17 years, were 0.90 (95% CI: 0.76–1.08), 0.84 (95% CI: 0.78–0.91), and 0.89 (95% CI: 0.84–0.95), respectively. Slightly stronger associations for income and education were seen among older women. Conclusions Lower levels of education and income were associated with modestly reduced fecundability. These results demonstrate the presence of socioeconomic disparities in fecundability.

Starting from historic reflections, the current SARS-CoV-2 induced COVID-19 pandemic is examined from various perspectives, in terms of what it implies for the implementation of non-pharmaceutical interventions, the modeling and monitoring of the epidemic, the development of early-warning systems, the study of mortality, prevalence estimation, diagnostic and serological testing, vaccine development, and ultimately clinical trials. Emphasis is placed on how the pandemic had led to unprecedented speed in methodological and clinical development, the pitfalls thereof, but also the opportunities that it engenders for national and international collaboration, and how it has simplified and sped up procedures. We also study the impact of the pandemic on clinical trials in other indications. We note that it has placed biostatistics, epidemiology, virology, infectiology, and vaccinology, and related fields in the spotlight in an unprecedented way, implying great opportunities, but also the need to communicate effectively, often amidst controversy. Buyse, M (corresponding author), Hasselt Univ, Interuniv Inst Biostat & Stat Bioinformat, Data Sci Inst, Martelarenlaan 42, Hasselt, Belgium. marc.buyse@iddi.com

Forecasting wastewater discharge is the basis for wastewater treatment and policy formulation. This paper proposes a novel mixed-data sampling regression model, i.e., combination-MIDAS model to forecast quarterly wastewater emissions in China based on dynamic factors at different frequencies. The results show that a significant auto-correlation for wastewater emissions exists and that water consumption per ten thousand gross domestic product is the best predictor of wastewater emissions. The forecast performances of the combination-MIDAS models are robust and better than those of the benchmark models. Therefore, the combination-MIDAS models can better capture the characteristics of wastewater emissions, suggesting that the proposed method is a good method to deal with model misspecification and uncertainty for the control and management of wastewater discharge in China. (C) 2019 Elsevier Ltd. All rights reserved.

Summary Small area ecological studies are commonly used in epidemiology to assess the impact of area level risk factors on health outcomes when data are only available in an aggregated form. However, the resulting estimates are often biased due to unmeasured confounders, which typically are not available from the standard administrative registries used for these studies. Extra information on confounders can be provided through external data sets such as surveys or cohorts, where the data are available at the individual level rather than at the area level; however, such data typically lack the geographical coverage of administrative registries. We develop a framework of analysis which combines ecological and individual level data from different sources to provide an adjusted estimate of area level risk factors which is less biased. Our method (i) summarizes all available individual level confounders into an area level scalar variable, which we call ecological propensity score (EPS), (ii) implements a hierarchical structured approach to impute the values of EPS whenever they are missing, and (iii) includes the estimated and imputed EPS into the ecological regression linking the risk factors to the health outcome. Through a simulation study, we show that integrating individual level data into small area analyses via EPS is a promising method to reduce the bias intrinsic in ecological studies due to unmeasured confounders; we also apply the method to a real case study to evaluate the effect of air pollution on coronary heart disease hospital admissions in Greater London.

Importance: Older and/or frail patients are underrepresented in landmark cancer trials. Tailored research is needed to address this evidence gap. Objective: The GO2 randomized clinical trial sought to optimize chemotherapy dosing in older and/or frail patients with advanced gastroesophageal cancer, and explored baseline geriatric assessment (GA) as a tool for treatment decision-making. Design, Setting, and Participants: This multicenter, noninferiority, open-label randomized trial took place at oncology clinics in the United Kingdom with nurse-led geriatric health assessment. Patients were recruited for whom full-dose combination chemotherapy was considered unsuitable because of advanced age and/or frailty. Interventions: There were 2 randomizations that were performed: CHEMO-INTENSITY compared oxaliplatin/capecitabine at Level A (oxaliplatin 130 mg/m 2on day 1, capecitabine 625 mg/m 2twice daily on days 1-21, on a 21-day cycle), Level B (doses 0.8 times A), or Level C (doses 0.6 times A). Alternatively, if the patient and clinician agreed the indication for chemotherapy was uncertain, the patient could instead enter CHEMO-BSC, comparing Level C vs best supportive care. Main Outcomes and Measures: First, broad noninferiority of the lower doses vs reference (Level A) was assessed using a permissive boundary of 34 days reduction in progression-free survival (PFS) (hazard ratio, HR = 1.34), selected as acceptable by a forum of patients and clinicians. Then, the patient experience was compared using Overall Treatment Utility (OTU), which combines efficacy, toxic effects, quality of life, and patient value/acceptability. For CHEMO-BSC, the main outcome measure was overall survival. Results: A total of 514 patients entered CHEMO-INTENSITY, of whom 385 (75%) were men and 299 (58%) were severely frail, with median age 76 years. Noninferior PFS was confirmed for Levels B vs A (HR = 1.09 [95% CI, 0.89-1.32]) and C vs A (HR = 1.10 [95% CI, 0.90-1.33]). Level C produced less toxic effects and better OTU than A or B. No subgroup benefited from higher doses: Level C produced better OTU even in younger or less frail patients. A total of 45 patients entered the CHEMO-BSC randomization: overall survival was nonsignificantly longer with chemotherapy: median 6.1 vs 3.0 months (HR = 0.69 [95% CI, 0.32-1.48], P =.34). In multivariate analysis in 522 patients with all variables available, baseline frailty, quality of life, and neutrophil to lymphocyte ratio were independently associated with OTU, and can be combined in a model to estimate the probability of different outcomes. Conclusions and Relevance: This phase 3 randomized clinical trial found that reduced-intensity chemotherapy provided a better patient experience without significantly compromising cancer control and should be considered for older and/or frail patients. Baseline geriatric assessment can help predict the utility of chemotherapy but did not identify a group benefiting from higher-dose treatment. Trial Registration: isrctn.org Identifier: ISRCTN44687907.

Economic panel data often exhibit cross-sectional dependence, even after conditioning on appropriate explanatory variables. Two approaches to modeling cross-sectional dependence in economic panel data are often used: the spatial dependence approach, which explains cross-sectional dependence in terms of distance among units, and the residual multifactor approach, which explains cross-sectional dependence by common factors that affect individuals to a different extent. This article reviews the theory on estimation and statistical inference for stationary and nonstationary panel data with cross-sectional dependence, particularly for models with a multifactor error structure. Tests and diagnostics for testing for unit roots, slope homogeneity, cointegration, and the number of factors are provided. We discuss issues such as estimating common factors, dealing with parameter plethora in practice, testing for structural stability and nonlinearity, and dealing with model and parameter uncertainty. Finally, we address issues related to the use of these economic panel models.

The central aim in this paper is to address variable selection questions in nonlinear and nonparametric regression. Motivated by statistical genetics, where nonlinear interactions are of particular interest, we introduce a novel and interpretable way to summarize the relative importance of predictor variables. Methodologically, we develop the "RelATive cEntrality" (RATE) measure to prioritize candidate genetic variants that are not just marginally important, but whose associations also stem from significant covarying relationships with other variants in the data. We illustrate RATE through Bayesian Gaussian process regression, but the methodological innovations apply to other "black box" methods. It is known that nonlinear models often exhibit greater predictive accuracy than linear models, particularly for phenotypes generated by complex genetic architectures. With detailed simulations and two real data association mapping studies, we show that applying RATE enables an explanation for this improved performance., 28 pages, 5 figures, 1 tables; Supplementary Material

Subject-specific and marginal models have been developed for the analysis of longitudinal ordinal data. Subject-specific models often lack a population-average interpretation of the model parameters due to the conditional formulation of random intercepts and slopes. Marginal models frequently lack an underlying distribution for ordinal data, in particular when generalized estimating equations are applied. To overcome these issues, latent variable models underneath the ordinal outcomes with a multivariate logistic distribution can be applied. In this article, we extend the work of O'Brien and Dunson (2004), who studied the multivariate t-distribution with marginal logistic distributions. We use maximum likelihood, instead of a Bayesian approach, and incorporated covariates in the correlation structure, in addition to the mean model. We compared our method with GEE and demonstrated that it performs better than GEE with respect to the fixed effect parameter estimation when the latent variables have an approximately elliptical distribution, and at least as good as GEE for other types of latent variable distributions.

Accepted manuscript version. Published version available at https://doi.org/10.1038/s41430-018-0271-2. Background: Epidemiological studies suggest that haem iron, which is found predominantly in red meat and increases endogenous formation of carcinogenic N-nitroso compounds, may be positively associated with lung cancer. The objective was to examine the relationship between haem iron intake and lung cancer risk using detailed smoking history data and serum cotinine to control for potential confounding. Methods: In the European Prospective Investigation into Cancer and Nutrition (EPIC), 416,746 individuals from 10 countries completed demographic and dietary questionnaires at recruitment. Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for incident lung cancer (n = 3731) risk relative to haem iron, non-haem iron, and total dietary iron intake. A corresponding analysis was conducted among a nested subset of 800 lung cancer cases and 1489 matched controls for whom serum cotinine was available. Results: Haem iron was associated with lung cancer risk, including after adjustment for details of smoking history (time since quitting, number of cigarettes per day): as a continuous variable (HR per 0.3 mg/1000 kcal 1.03, 95% CI 1.00–1.07), and in the highest versus lowest quintile (HR 1.16, 95% CI 1.02–1.32; trend across quintiles: P = 0.035). In contrast, non-haem iron intake was related inversely with lung cancer risk; however, this association attenuated after adjustment for smoking history. Additional adjustment for serum cotinine did not considerably alter the associations detected in the nested case–control subset. Conclusions: Greater haem iron intake may be modestly associated with lung cancer risk.

With a rapid increase in volume and complexity of data sets, there is a need for methods that can extract useful information, for example the relationship between two data sets measured for the same persons. The Partial Least Squares (PLS) method can be used for this dimension reduction task. Within life sciences, results across studies are compared and combined. Therefore, parameters need to be identifiable, which is not the case for PLS. In addition, PLS is an algorithm, while epidemiological study designs are often outcome-dependent and methods to analyze such data require a probabilistic formulation. Moreover, a probabilistic model provides a statistical framework for inference. To address these issues, we develop Probabilistic PLS (PPLS). We derive maximum likelihood estimators that satisfy the identifiability conditions by using an EM algorithm with a constrained optimization in the M step. We show that the PPLS parameters are identifiable up to sign. A simulation study is conducted to study the performance of PPLS compared to existing methods. The PPLS estimates performed well in various scenarios, even in high dimensions. Most notably, the estimates seem to be robust against departures from normality. To illustrate our method, we applied it to IgG glycan data from two cohorts. Our PPLS model provided insight as well as interpretable results across the two cohorts., Comment: Accepted in Journal of Multivariate Analysis

BACKGROUND: IDeAl (Integrated designs and analysis of small population clinical trials) is an EU funded project developing new statistical design and analysis methodologies for clinical trials in small population groups. Here we provide an overview of IDeAl findings and give recommendations to applied researchers. METHOD: The description of the findings is broken down by the nine scientific IDeAl work packages and summarizes results from the project's more than 60 publications to date in peer reviewed journals. In addition, we applied text mining to evaluate the publications and the IDeAl work packages' output in relation to the design and analysis terms derived from in the IRDiRC task force report on small population clinical trials. RESULTS: The results are summarized, describing the developments from an applied viewpoint. The main result presented here are 33 practical recommendations drawn from the work, giving researchers a comprehensive guidance to the improved methodology. In particular, the findings will help design and analyse efficient clinical trials in rare diseases with limited number of patients available. We developed a network representation relating the hot topics developed by the IRDiRC task force on small population clinical trials to IDeAl's work as well as relating important methodologies by IDeAl's definition necessary to consider in design and analysis of small-population clinical trials. These network representation establish a new perspective on design and analysis of small-population clinical trials. CONCLUSION: IDeAl has provided a huge number of options to refine the statistical methodology for small-population clinical trials from various perspectives. A total of 33 recommendations developed and related to the work packages help the researcher to design small population clinical trial. The route to improvements is displayed in IDeAl-network representing important statistical methodological skills necessary to design and analysis of small-population clinical trials. The methods are ready for use. ispartof: ORPHANET JOURNAL OF RARE DISEASES vol:13 issue:1 ispartof: location:England status: published

The relationship between smoking and suicide remains controversial. A total of 16 282 twin pairs born before 1958 in Finland and alive in 1974 were queried with detailed health and smoking questionnaires in 1975 and 1981, with response rates of 89% and 84%. Smoking status and dose, marital, employment, and socio-economic status, and indicators of psychiatric and somatic illness were assessed at both time points. Emergent psychiatric and medical illness and vital status, including suicide determined by forensic autopsy, were evaluated over 35-year follow-up through government registries. The association between smoking and suicide was determined in competing risks hazard models. In twin pairs discordant for smoking and suicide, the prospective association between smoking and suicide was determined using a matched case–control design. Smokers had a higher cumulative suicide incidence than former or never smokers. Heavy smokers had significantly higher suicide risk [hazard ratio (HR) 3.47, 95% confidence interval (CI) 2.31–5.22] than light smokers (HR 2.30, 95% CI 1.61–3.23) (p = 0.017). Compared with never smokers, smokers, but not former smokers, had increased suicide risk (HR 2.56, 95% CI 1.43–4.59), adjusting for depressive symptoms, alcohol and sedative–hypnotic use, and excluding those who developed serious somatic or psychiatric illness. In twin pairs discordant for smoking and suicide, suicide was more likely in smokers [odds ratio (OR) 6.0, 95% CI 2.06–23.8]. Adults who smoked tobacco were more likely to die by suicide, with a large, dose-dependent effect. This effect remained after consideration of many known predictors of suicide and shared familial effects, consistent with the hypothesis that exposure to tobacco smoke increases the risk of suicide., published version, peerReviewed

The selection of the optimal number of components remains a difficult but essential task in partial least squares (PLS). Randomization tests have the advantage of being automatic and they make use of the entire dataset, in contrary with the widely used cross-validation approaches. Partial least squares modeling may include component(s) with a large amount of irrelevant data variation, and this might affect the model, depending on the assigned y-loading (which is the regression coefficient in the latent domain). This has recently been indicated by us in the basic sequence framework with respect to the underlying theory of the PLS algorithm and presented to the chemometrics society. We will show in this work that this irrelevant data variation is the root cause of the difficulty in current methods for selecting the optimal number of components. For randomization tests, PLS models with nonsignificant components may result in false positive tests because of the incorrect assumption that "the components enter the model in a natural order". In this work, we introduce a new randomization test, weight randomization test, selection of the optimal number of components in PLS in light of the underlying theory of the PLS algorithm. In the proposed method the null distribution is well characterized and efficiently determined taking into account a newly defined model quality metric: the number of consecutive non-significant components (CNC). We illustrate the effectiveness of weight randomization test in optimization of preprocessing as well as in classification models, where results are compared with the double cross-validation procedure for the latter. This is an important step towards the full automation of PLS model development and routine updates.

This article studies design selection for generalized linear models (GLMs) using the quantile dispersion graphs (QDGs) approach in the presence of misspecification in the link and/or linear predictor. The uncertainty in the linear predictor is represented by a unknown function and estimated using kriging. For addressing misspecified link functions, a generalized family of link functions is used. Numerical examples are shown to illustrate the proposed methodology.

The efficient bootstrap methodology is developed for overidentified moment conditions models with weakly dependent observation. The resulting bootstrap procedure is shown to be asymptotically valid and can be used to approximate the distributions of t-statistics, the J-statistic for overidentifying restrictions, and Wald, Lagrange multiplier and distance statistics for nonlinear hypotheses. The asymptotic validity of the efficient bootstrap based on a computationally less demanding approximate k-step estimator is also shown. The finite sample performance of the proposed bootstrap is assessed using simulations in an intertemporal consumption based asset pricing model. (C) 2010 Elsevier B.V. All rights reserved.

Contains fulltext : 103199.pdf (Publisher’s version ) (Open Access) In this paper, we compare by means of Monte Carlo simulations two approaches to take spatial autocorrelation into account: the classical spatial autoregressive model and the structural equations model with latent variables. The former accounts for spatial dependence and spillover effects in georeferenced data by means of a spatial weights matrix W. The latter represents spatial dependence and spillover effects by means of a latent variable in the structural (regression) model while the observed spatially lagged variables are related to the latent spatial dependence variable in the measurement model. The simulation results based on Anselin's Columbus, Ohio, crime data set show that the misspecified latent variables approach slightly trails the correctly specified classical approach in terms of bias and root mean squared error of the coefficient estimators. 21 p.

Aim To investigate nurse managers' leadership behaviour in relation to the sickness absence records of nursing staff.Background Sickness absence is high in healthcare and interferes with nursing efficiency and quality. Nurse managers' leadership behaviour may be associated with nursing staff sickness absence.Method Six nurse managers completed the Leadership Effectiveness and Adaptability Description (LEAD) questionnaire, which assesses leadership behaviour in terms of leadership flexibility (i.e. the range of leadership styles) and effectiveness (i.e. using the leadership style that is appropriate for a given situation). LEAD scores were linked to the number of recorded days of sickness absence and both short (17 days) and long (>7 days) episodes of sickness absence in the nursing teams.Results Leadership flexibility of nurse managers was not associated with sickness absence among nurses. High leadership effectiveness was associated with fewer days and fewer short episodes of sickness absence. Leadership effectiveness was unrelated to the number of long episodes of sickness absence.Conclusion Effective nurse managers had less short-term sickness absence in their nursing teams.Implications for nursing management If these tentative cross-sectional associations are confirmed in longitudinal studies including more departments, then training effective leadership may improve the management of short-term sickness absence.

We show that variation in an intronic length polymorphism in the CHD1-Z gene in Black-tailed Godwits Limosa l. limosa is associated with fitness correlates. This is the second example of the CHDZ-1 gene being correlated with fitness, a previous study having established that Moorhens Gallinula chloropus carrying the rare Z* allele have reduced survival. In Godwits, however, carriers of the Z* allele (374 bp) fared better than those with the more frequent Z allele (378 bp) with respect to body mass, plumage ornamentation, reproductive parameters and habitat quality. The Z* allele was found in 14% of 251 adult birds from nature reserves, but was absent from 33 birds breeding in intensively managed agricultural lands. Males and females with the Z* allele had less extensive breeding plumage, and females had a higher body mass, bred earlier and had larger eggs. There were no significant differences in annual survival between birds with and without the Z* allele. DNA isolated from museum skins demonstrated that this polymorphism was present at low frequency in 1929. We speculate that strong asymmetrical overdominance may explain the low frequency of the Z* allele and that genetic linkage to causal genes might be an explanation for the phenotypic correlations. Our findings suggest a degree of cryptic genetic population structuring in the Dutch Godwit population.

This article reviews inverse probability weighting methods and doubly robust estimation methods for the analysis of incomplete data sets. We first consider methods for estimating a population mean when the outcome is missing at random, in the sense that measured covariates can explain whether or not the outcome is observed. We then sketch the rationale of these methods and elaborate on their usefulness in the presence of influential inverse weights. We finally outline how to apply these methods in a variety of settings, such as for fitting regression models with incomplete outcomes or covariates, emphasizing the use of standard software programs.

In spite of the strict EU regulations, concentrations of surface ozone and PM10 often exceed the pollution standards for the Netherlands and Europe. Their concentrations are controlled by (precursor) emissions, social and economic developments and a complex combination of meteorological actors. This study tackles the latter, and provides insight in the meteorological processes that play a role in O3 and PM10 levels in rural mid-latitudes sites in the Netherlands. The relations between meteorological actors and air quality are studied on a local scale based on observations from four rural sites and are determined by a comprehensive correlation analysis and a multiple regression (MLR) analysis in 2 modes, with and without air quality variables as predictors. Furthermore, the objective Lamb Weather Type approach is used to assess the influence of the large-scale circulation on air quality. Keeping in mind its future use in downscaling future climate scenarios for air quality purposes, special emphasis is given to an appropriate selection of the regressor variables readily available from operational meteorological forecasts or AOGCMs (Atmosphere-Ocean coupled General Circulation Models). The regression models perform satisfactory, especially for O3, with an (R2 of 57.0% and 25.0% for PM10. Including previous day air quality information increases significantly the models performance by 15% (O3) and 18% (PM10). The Lamb weather types show a seasonal distinct pattern for high (low) episodes of average O3 and PM10 concentrations, and these are clear related with the meteorology-air quality correlation analysis. Although using a circulation type approach can provide important additional physical relations forward, our analysis reveals the circulation method is limited in terms of short-term air quality forecast for both O3 and PM10 (R2 between 0.12 and 23%). In summary, it is concluded that the use of a regression model is more promising for short-term downscaling from climate scenarios than the use of a weather type classification approach.

Understanding how environmental factors could impact population dynamics is of primary importance for species conservation. Matrix population models are widely used to predict population dynamics. However, in species-rich ecosystems with many rare species, the small population sizes hinder a good fit of species-specific models. In addition, classical matrix models do not take into account environmental variability. We propose a mixture of regression models with variable selection allowing the simultaneous clustering of species into groups according to vital rate information (recruitment, growth and mortality) and the identification of group-specific explicative environmental variables. We develop an inference method coupling the R packages flexmix and glmnet. We first highlight the effectiveness of the method on simulated datasets. Next, we apply it to data from a tropical rain forest in the Central African Republic. We demonstrate the accuracy of the inhomogeneous mixture matrix model in successfully reproducing stand dynamics and classifying tree species into well-differentiated groups with clear ecological interpretations. Copyright (c) 2014 John Wiley & Sons, Ltd.

Objective To develop and validate the Italian Health Search Morbidity (HSM) Index to adjust health care costs in general practice. Methods The study population comprised 1,076,311 patients registered in the Health Search CSD Longitudinal Patient Database between January 1, 2008, and December 31, 2010. We randomly selected 538,254 and 538,057 patients to form the development and validation cohorts, respectively. To ensure model convergence, 5% of the aforementioned cohorts were selected randomly to create development and validation samples. The outcome was the total direct health care costs covered by the national health system. Interaction between age and sex, chronic diseases, and acute diseases were entered in a multilevel generalized linear latent mixed model with random intercepts (province of residence and general practitioner) to identify determinants associated with increased or decreased costs. The estimated coefficients were linearly combined to create the HSM Index for individual patients. The score was applied to the validation sample, and measures of predictive accuracy, explained variance, and the observed/predicted ratio were computed to evaluate the model's accuracy. Results The mean yearly cost was €414.57 per patient, and the HSM Index had a median value of 5.08 (25th–75th range 4.44–5.98). The HSM Index explained 50.17% of the variation in costs. Concerning calibration, in 80% of the population, the margin of error in the estimation of costs was around 10%. Conclusions The HSM Index is a reliable case-mix system that could be implemented in general practice for costs adjustment. This tool should ensure fairer scrutiny of resource use and allocation of budgets among general practitioners.

© 2017 ISI/BS. In this paper, we consider a semiparametric promotion time cure model and study the asymptotic properties of its nonparametric maximum likelihood estimator (NPMLE). First, by relying on a profile likelihood approach, we show that the NPMLE may be computed by a single maximization over a set whose dimension equals the dimension of the covariates plus one. Next, using Z-estimation theory for semiparametric models, we derive the asymptotics of both the parametric and nonparametric components of the model and show their efficiency. We also express the asymptotic variance of the estimator of the parametric component. Since the variance is difficult to estimate, we develop a weighted bootstrap procedure that allows for a consistent approximation of the asymptotic law of the estimators. As in the Cox model, it turns out that suitable tools are the martingale theory for counting processes and the infinite dimensional Z-estimation theory. Finally, by means of simulations, we show the accuracy of the bootstrap approximation. ispartof: BERNOULLI vol:23 issue:4B pages:3437-3468 status: published

An important characteristic of chemometrics has been its need to manage the tradeoff between computational, mathematical and statistical performance against data interpretability. Additionally, being mostly seen as a conglomeration of data analytic methods that target the solution to real-world problems, the development of chemometrics as an independent and well-defined field has been hampered by its applied nature. Consequently, the broad range and diversity of application of chemometric tools has hindered the development of a unified theory able to propel it beyond its current use in analytical and industrial chemistry to larger and more complex data problems.In this paper, we provide a mathematical vehicle for the understanding and improvement of current methods popular in chemometrics. Starting from a historical solution to matrix factorization we develop a novel unified framework for the fundamentals of latent variable modeling methods, elucidate major properties and clarify controversies between major PLS implementations and interpretations. The concepts presented in this work aims at contributing to a deeper understanding of the underlying theory of chemometrics methods, and strengthen their use in practice. Furthermore, this effort attempts to bridge the gap between chemometrics and big data problems and contribute to the development and acceptance of chemometrics as a mature and independent scientific field by the broader data analytic community. (C) 2015 Elsevier B.V. All rights reserved.

Background: The axillary nodal status may influence the prognosis and the choice of adjuvant treatment of individual breast cancer patients. The variation in number of reported axillary lymph nodes and its effect on the axillary nodal stage were studied and the implications are discussed.Methods: Between 1994 and 1997, a total of 4,806 axillary dissections for invasive breast cancers in 4,715 patients were performed in hospitals in the North-Netherlands. The factors associated with the number of reported nodes and the relation of this number with the nodal status and the number of positive nodes were studied.Results: The number of reported nodes varied significantly between pathology laboratories, the median number of nodes ranged from 9 to 15, respectively. The individual hospitals explained even more variability in the number of nodes than pathology laboratories (range in median number 8-15, P 20 nodes were examined, the percentage of tumors with >4 positive nodes increased from 4 to 31%. Multivariate analysis confirmed these results.Conclusions: This population-based study showed a large variation in the number of reported lymph nodes between hospitals. A more extensive surgical dissection or histopathological examination of the specimen generally resulted in a higher number of positive nodes. Although the impact of misclassification on adjuvant treatment will have varied, the impact with regard to adjuvant regional radiotherapy may have been considerable. (C) 2004 Wiley-Liss, Inc.

Pancreatic cancer has few early symptoms, is usually diagnosed at late stages, and has a high case-fatality rate. Identifying modifiable risk factors is crucial to reducing pancreatic cancer morbidity and mortality. Prior studies have suggested that specific foods and nutrients, such as dairy products and constituents, may play a role in pancreatic carcinogenesis. In this pooled analysis of the primary data from 14 prospective cohort studies, 2212 incident pancreatic cancer cases were identified during follow-up among 862 680 individuals. Adjusting for smoking habits, personal history of diabetes, alcohol intake, body mass index (BMI), and energy intake, multivariable study-specific hazard ratios (MVHR) and 95% confidence intervals (CIs) were calculated using the Cox proportional hazards models and then pooled using a random effects model. There was no association between total milk intake and pancreatic cancer risk (MVHR = 0.98, 95% CI = 0.82-1.18 comparing ≥500 with 1-69.9 g/day). Similarly, intakes of low-fat milk, whole milk, cheese, cottage cheese, yogurt, and icecream were not associated with pancreatic cancer risk. No statistically significant association was observed between dietary (MVHR = 0.96, 95% CI = 0.77-1.19) and total calcium (MVHR = 0.89, 95% CI = 0.71-1.12) intake and pancreatic cancer risk overall when comparing intakes ≥1300 with

A systematic approach is described for building validated PLS models that predict cholesterol and triglyceride concentrations in lipoprotein subclasses in fasting serum from a normolipidemic, healthy population. The PLS models were built on diffusion-edited 1H NMR spectra and calibrated on HPLC-derived lipoprotein subclasses. The PLS models were validated using an independent test set. In addition to total VLDL, LDL, and HDL lipoproteins, statistically significant PLS models were obtained for 13 subclasses, including 5 VLDLs (particle size 64-31.3 nm), 4 LDLs (particle size 28.6-20.7 nm) and 4 HDLs (particle size 13.5-9.8 nm). The best models were obtained for triglycerides in VLDL (0.82 < Q2

Extreme events such as heat waves, cold spells, floods, droughts, tropical cyclones, and tornadoes have potentially devastating impacts on natural and engineered systems and human communities worldwide. Stakeholder decisions about critical infrastructures, natural resources, emergency preparedness and humanitarian aid typically need to be made at local to regional scales over seasonal to decadal planning horizons. However, credible climate change attribution and reliable projections at more localized and shorter time scales remain grand challenges. Long-standing gaps include inadequate understanding of processes such as cloud physics and ocean–land–atmosphere interactions, limitations of physics-based computer models, and the importance of intrinsic climate system variability at decadal horizons. Meanwhile, the growing size and complexity of climate data from model simulations and remote sensors increases opportunities to address these scientific gaps. This perspectives article explores the possibility that physically cognizant mining of massive climate data may lead to significant advances in generating credible predictive insights about climate extremes and in turn translating them to actionable metrics and information for adaptation and policy. Specifically, we propose that data mining techniques geared towards extremes can help tackle the grand challenges in the development of interpretable climate projections, predictability, and uncertainty assessments. To be successful, scalable methods will need to handle what has been called "big data" to tease out elusive but robust statistics of extremes and change from what is ultimately small data. Physically based relationships (where available) and conceptual understanding (where appropriate) are needed to guide methods development and interpretation of results. Such approaches may be especially relevant in situations where computer models may not be able to fully encapsulate current process understanding, yet the wealth of data may offer additional insights. Large-scale interdisciplinary team efforts, involving domain experts and individual researchers who span disciplines, will be necessary to address the challenge.

Studies in epidemiology and social sciences are often longitudinal and outcome measures are frequently obtained by questionnaires in ordinal scales. To understand the relationship between explanatory variables and outcome measures, generalized estimating equations can be applied to provide a population-averaged interpretation and address the correlation between outcome measures. It can be performed by different software packages, but a motivating example showed differences in the output. This paper investigated the performance of GEE in R (version 3.0.2), SAS (version 9.4), and SPSS (version 22.0.0) using simulated data under default settings. Multivariate logistic distributions were used in the simulation to generate correlated ordinal data. The simulation study demonstrated substantial bias in the parameter estimates and numerical issues for data sets with relative small number of subjects. The unstructured working association matrix requires larger numbers of subjects than the independence and exchangeable working association matrices to reduce the bias and diminish numerical issues. The coverage probabilities of the confidence intervals for fixed parameters were satisfactory for the independence and exchangeable working association matrix, but they were frequently liberal for the unstructured option. Based on the performance and the available options, SPSS and multgee, and repolr in R all perform quite well for relatively large sample sizes (e.g. 300 subjects), but multgee seems to do a little better than SPSS and repolr in most settings. (C) 2014 Elsevier B.V. All rights reserved.

This paper presents a first-order autoregressive distributed lag model in both space and time. It is shown that this model encompasses a wide series of simpler models frequently used in the analysis of space-time data as well as models that better fit the data and have never been used before. A framework is developed to determine which model is the most likely candidate to study space-time data. As an application, the relationship between the labor force participation rate and the unemployment rate is estimated using regional data of Germany, France, and the United Kingdom derived from Eurostat, 1983-1993.

We propose a method for fitting semiparametric models such as the proportional hazards (PH), additive risks (AR), and proportional odds (PO) models. Each of these semiparametric models implies that some transformation of the conditional cumulative hazard function (at each t) depends linearly on the covariates. The proposed method is based on nonparametric estimation of the conditional cumulative hazard function, forming a weighted average over a range of t-values, and subsequent use of least squares to estimate the parameters suggested by each model. An approximation to the optimal weight function is given. This allows semiparametric models to be fitted even in incomplete data cases where the partial likelihood fails (e.g., left censoring, right truncation). However, the main advantage of this method rests in the fact that neither the interpretation of the parameters nor the validity of the analysis depend on the appropriateness of the PH or any of the other semiparametric models. In fact, we propose an integrated method for data analysis where the role of the various semiparametric models is to suggest the best fitting transformation. A single continuous covariate and several categorical covariates (factors) are allowed. Simulation studies indicate that the test statistics and confidence intervals have good small-sample performance. A real data set is analyzed.

We have introduced multilinear PLS in 3D QSAR and applied it to GRID descriptors from a set of benzamides with affinity to the dopamine D-3 receptor subtype, synthesized as potential drugs against schizophrenia. The key issue in 3D QSAR modelling is to obtain a predictive model that is easy to interpret, Each component in the multilinear PLS model explains clearly defined details, e.g. substituent positions, while the bilinear PLS solution is general and more difficult to interpret. The best models were obtained after four components with multilinear PLS (Q(2) = 51%) and after only one component with bilinear PLS (Q(2) = 50%). The external test set was predicted better with multilinear PLS (Q(2) = 31%) than with bilinear PLS (Q(2) = 25%). With multilinear PLS one loses in fit and gains in stability and simplicity owing to the fewer parameters that need to be estimated as compared with bilinear PLS. Finally, multilinear PLS is also less influenced by insignificant variation in the descriptor block, which is an advantage in 3D QSAR modelling. (C) 1997 John Wiley & Sons, Ltd.

Background Exposure to air pollution is frequently associated with reductions in birth weight but results of available studies vary widely, possibly in part because of differences in air pollution metrics. Further insight is needed to identify the air pollution metrics most strongly and consistently associated with birth weight. Methods We used a hospital-based obstetric database of more than 70,000 births to study the relationships between air pollution and the risk of low birth weight (LBW, Results Increased risks of LBW were associated with ambient O3 concentrations as measured by monitoring stations, as well as traffic density and proximity to major roadways. LBW was not significantly associated with other air pollution metrics, except that a decreased risk was associated with ambient NO2 concentrations as measured by monitoring stations. When birth weight was analyzed as a continuous variable, small increases in mean birth weight were associated with most air pollution metrics (3 concentrations. Conclusions We found contrasting results according to the different air pollution metrics examined. Unmeasured confounders and/or measurement errors might have produced spurious positive associations between birth weight and some air pollution metrics. Despite this, ambient O3 was associated with a decrement in mean birth weight and significant increases in the risk of LBW were associated with traffic density, proximity to roads and ambient O3. This suggests that in our study population, these air pollution metrics are more likely related to increased risks of LBW than the other metrics we studied. Further studies are necessary to assess the consistency of such patterns across populations.

Recently, Schroeder etal. (2010, Ibis 152: 368-377) suggested that intronic variation in the CHD1-Z gene of Black-tailed Godwits breeding in southwest Friesland, The Netherlands, correlated with fitness components. Here we re-examine this surprising result using an expanded dataset (2088 birds sampled from 2004 to 2010 vs. 284 birds from 2004 to 2007). We find that the presence of the Z* allele (9% of the birds) is not associated with breeding habitat type, egg size, adult survival, adult body mass or adult body condition. The results presented here, when used in synergy with the previously reported results by Schroeder etal., suggest that there might be a tendency towards female adults with the Z* allele laying earlier clutches than adult females without the Z* allele. The occurrence of the Z* allele was also associated with a higher chick body mass and return rate. Chicks with the Z* allele that had hatched early in the breeding season were heavier at birth than chicks without the Z* allele and chicks with the Z* allele that had hatched late. Collectively, the results suggest that variation in the CHD1-Z gene may indeed have arisen as a byproduct of selection acting on females during the egg fase and on chicks during the rearing stages of the reproductive cycle.

In recent years, there has been increasing interest in modelling of species abundance data in addition to presence data. In this study, we assessed the similarities and differences between presence-absence distributions and abundance distributions along similar environmental gradients, derived, respectively, from presence-absence and abundance data. Moreover, we examined the possibility of using presence-absence distribution models to derive abundance distributions. For this purpose, we used Braun-Blanquet abundance scores for 243 vascular species at 10 996 French forest sites. Species distribution models were used to analyse the link between the patterns of occurrence, low abundance and high abundance for each species with regard to mean annual temperature, June water balance, and soil pH. For each species, differences in the modelled distributions were characterised by the ecological optimum and ecological amplitude. A comparison of the presence-absence and abundance distributions for all species revealed similar optima and different amplitudes along the three ecological factors. An abundant-centre distribution was observed in environmental space, with species abundance being greatest at the optimal conditions and lower at less favourable conditions of the species occurrence response. Geographical habitat mapping also shows centred, high-abundance suitability within the presence habitat of each species. We conclude that species distribution models derived from presence-absence data provide useful information about the ecological optima of abundance distributions but overestimate the range of habitats suitable for high species abundance. This study demonstrates the utility of presence-absence data for ecologist and conservation biologist when they are interested in the optimal conditions of high species abundance.

The purpose of this article is to demonstrate the use of the quantile dispersion graphs (QDGs) approach for comparing candidate designs for generalized linear models in the presence of model misspecification in the linear predictor. The proposed design criterion is based on the mean-squared error of prediction which incorporates the prediction variance and the bias caused by fitting the wrong model. The method of kriging is used to estimate the unknown function assumed to be the cause of model misspecification. The QDGs approach is also useful in assessing the robustness of a given design to values of the unknown parameters in the linear predictor. Three numerical examples are presented to illustrate the application of the proposed methodology. (C) 2011 Elsevier B.V. All rights reserved.

This article proposes a family of link functions for the multinomial response model. The link family includes the multicategorical logistic link as one of its members. Conditions for the local orthogonality of the link and the regression parameters are given. It is shown that local orthogonality of the parameters in a neighbourhood makes the link family location and scale invariant. Confidence regions for jointly estimating the percentiles based on the parametric family of link functions are also determined. A numerical example based on a combination drug study is used to illustrate the proposed parametric link family and the confidence regions for joint percentile estimation., Comment: 28 pages, 4 tables and 1 figure

This paper addresses the issue of Software Cost Estimation (SCE) providing an alternative approach to modelling and prediction using Artificial Neural Networks (ANN) and Input Sensitivity Analysis (ISA). The overall aim is to identify and investigate the effect of the leading factors in SCE, through ISA. The factors identified decisively influence software effort in the models examined and their ability to provide sufficiently accurate SCEs is examined. ANN of variable topologies are trained to predict effort devoted to software development based on past (finished) projects recorded in two publicly available historical datasets. The main difference with relevant studies is that the proposed approach extracts the most influential cost drivers that describe best the effort devoted to development activities using the weights of the network connections. The approach is validated on known software cost data and the results obtained are assessed and compared. The ANN constructed generalise efficiently the knowledge acquired during training providing accurate effort predictions. The validation process included predictions with only the most highly ranked attributes among the original cost attributes of the datasets and revealed that accuracy performance was maintained at same levels. The results showed that the combination of ANN and ISA is an effective method for evaluating the contribution of cost factors, whereas the subsets of factors selected did not compromise the accuracy of the prediction results.