Shrinkage Procedure Research Articles

Robust day-ahead solar forecasting with endogenous data and sliding windows

Renewable energy forecasting services comprise various modules for intra-day and day-ahead forecasts. This work specifically addresses day-ahead forecasts, utilizing specifications based on endogenous, historical measurements. These specifications are designed to be computationally efficient, requiring fewer input variables and less training data. Such weather-independent specifications serve as benchmarks against the more computationally demanding forecasts based on numerical weather predictions. A series of experiments, designed to simulate the real-world application of an online system, were conducted on sliding windows of back-contact photovoltaic (installed at KAUST, Saudi Arabia) output series, solar irradiance recorded in Hawaii, and simulated data. Our analysis evaluated 24 specifications, which are variants of (i) functional time series models (including two novel shrinkage procedures); (ii) time series nearest neighbor schemes; (iii) exponential smoothing procedures; (iv) autoregressive integrated moving average processes; (v) automatic techniques based on time series decomposition; and (vi) the persistence model. In addition to employing outlier-robust accuracy metrics, such as mean absolute error, our evaluation also prioritized prediction-interval accuracy, quantified by the mean scaled interval score. Our findings suggest that practitioners can achieve significant improvements over the persistence model by forecasting daily profiles using adaptive nonparametric or functional data analysis-based procedures. Moreover, applying shrinkage to nearest neighbor (NN) forecasts toward smooth, average daily profiles significantly enhances NN performance. Conversely, some popular, computationally intensive models fail to perform adequately to justify their additional cost.

Sparse generalized Yule–Walker estimation for large spatio-temporal autoregressions with an application to NO2 satellite data

We consider a high-dimensional model in which variables are observed over time and space. The model consists of a spatio-temporal regression containing a time lag and a spatial lag of the dependent variable. Unlike classical spatial autoregressive models, we do not rely on a predetermined spatial interaction matrix, but infer all spatial interactions from the data. Assuming sparsity, we estimate the spatial and temporal dependence fully data-driven by penalizing a set of Yule–Walker equations. This regularization can be left unstructured, but we also propose customized shrinkage procedures when observations originate from spatial grids (e.g. satellite images). Finite sample error bounds are derived and estimation consistency is established in an asymptotic framework wherein the sample size and the number of spatial units diverge jointly. Exogenous variables can be included as well. A simulation exercise shows strong finite sample performance compared to competing procedures. As an empirical application, we model satellite measured nitrogen dioxide (NO2) concentrations in London. Our approach delivers forecast improvements over a competitive benchmark and we discover evidence for strong spatial interactions.

Model detection for semiparametric accelerated failure additive model with right-censored data.

Censored data frequently appeared in applications across a variety of different areas like epidemiology or medical research. Traditionally statistical inference on this data mechanism was based on some pre-assigned models that will suffer from the risk of model-misspecification. This article proposes a two-folded shrinkage procedure for simultaneous structure identification and variable selection of the semiparametric accelerated failure additive model with right-censored data, in which the nonparametric functions are addressed by spline approximation. Under some regularity conditions, the consistency of model structure identification is theoretically established in the sense that the proposed method can automatically separate the linear and zero components from the nonlinear ones with probability approaching to one. Detailed issues in computation and turning parameter selection are also discussed. Finally, we illustrate the proposed method by some simulation studies and two real data applications to the primary biliary cirrhosis data and skin cutaneous melanoma data.

Material Extrusion Additive Manufacturing of Poly(Lactic Acid): Influence of infill orientation angle

The effect that the infill orientation angle has on the strain-rate dependence of the yield stress for material extrusion additive manufactured (ME-AM) PolyLactic Acid (PLA) material was investigated. Symmetric angle-ply stacking sequences were used to produce ME-AM tensile test samples. Measured yield stresses were compensated for the voided structure, typical of ME-AM components. Furthermore, molecular orientation and stretch was macroscopically assessed by a thermal shrinkage procedure. Additionally, hot-press compression molded ( CM ) samples were manufactured and mechanically characterized in uniaxial tensile and compression in order to determine the material’s isotropic bulk properties. Initial model parameters for the Ree–Eyring modification of the Eyring flow rule were determined using CM data. According to SEM fractography, all samples showed microscopically brittle fracture behavior. Notwithstanding, contrary to CM samples, ME-AM specimens showed macroscopically ductile stress–strain behavior and a transition from a regime with only a primary α -deformation process, at low strain rates, to a regime with 2 deformation processes ( α + β ), at high strain rates. These effects are an influence of the processing step and are attributed to the molecular orientation and stretch of the polymer chains, provoking anisotropic mechanical properties. As a consequence, a deformation-induced change of the Eyring rate constants is needed to adequately describe the strain-rate dependence of the ME-AM yield stress behavior, leaving the initial activation volumes unchanged. Taking this deformation-dependence of the rate constants into account, yield stresses as a function of infill orientation angle can be appropriately predicted. • ME-AM samples with different symmetric angle-ply stacking sequences were produced. • PLA shows anisotropy in yield stress and strain-rate dependent behavior. • ME-AM samples show ductile stress–strain behavior, due to molecular orientation. • ME-AM samples show transition in strain-rate dependence, due to molecular orientation. • PLA demonstrates deformation-dependent Eyring rate constants.

Analysis of A Disintegrin and Metalloprotease 17 (ADAM17) Expression as a Prognostic Marker in Ovarian Cancer Patients Undergoing First-Line Treatment Plus Bevacizumab.

To find prognostic factors for advanced ovarian cancer patients undergoing first-line therapy with carboplatin, paclitaxel and bevacizumab, we investigated the expression of a disintegrin and metalloprotease 17 (ADAM17) in cancer tissues. ADAM17 has been involved in ovarian cancer development, progression and cell resistance to cisplatin. Tissue microarrays from 309 ovarian cancer patients enrolled in the MITO16A/MANGO-OV2 clinical trial were analyzed by immunohistochemistry for ADAM17 protein expression. Intensity and extent of staining were combined into a semi-quantitative visual grading system (H score) which was related to clinicopathological characteristics of cases and the clinical outcome of patients by univariate and multivariate Cox regression models. ADAM17 immunostaining was detected in most samples, mainly localized in the tumor cells, with variable intensity across the cohort. Kaplan–Meier survival curves, generated according to the best cut-off value for the ADAM17 H score, showed that high ADAM17 expression was associated with worse prognosis for PFS and OS. However, after the application of a shrinkage procedure to adjust for overfitting hazard ratio estimates, the ADAM17 value as prognostic factor was lost. As subgroup analysis suggested that ADAM17 expression could be prognostically relevant in cases with no residual disease at baseline, further studies in this patient category may be worth planning.

Role of Wrist Arthroscopy in Juvenile Inflammatory Arthritis.

Background and Purposes The wrist is the fourth most common joint to be involved in juvenile inflammatory arthritis (JIA), which is a common rheumatological condition affecting children. Wrist arthroscopy is well established in rheumatoid arthritis, but remains unexplored in JIA. The aim of this study is to investigate the role of wrist arthroscopy in JIA, with focus on those who are refractory to medical management. Methods This is a prospective observational study, including consecutive patients with JIA undergoing arthroscopy between January 2016 and December 2020. Those over the age of 18 years and those with other rheumatological diagnoses were excluded. Data including pre-, intra-, and postoperative variables, demographics, and patient-reported outcomes were collated and are reported using standard measures. Results A total of 15 patients underwent arthroscopy ( n = 20 wrists). Synovitis was noted in all wrists on arthroscopy and synovectomy was performed in all cases. Other procedures were performed as indicated during the procedure. The median follow-up duration was 11.3 (interquartile range [IQR] 8.1-24.2) months. Median reduction of 4 (IQR 2.25-6) points on the Visual Analogue Score for pain on loading was noted postoperatively. Grip strength was improved in n = 11/20 wrists and functional improvement was noted in n = 18/20 wrists. Restriction of range of motion was achieved with a shrinkage procedure in patients with hypermobile joints. There were no postoperative complications, and no patients were lost to follow-up. Conclusion In experienced hands, wrist arthroscopy is feasible, safe, and efficacious in the management of JIA, among patients who are refractory to medical management. Level of Evidence This is a Level II study.

Evaluation of Angiogenesis-Related Genes as Prognostic Biomarkers of Bevacizumab Treated Ovarian Cancer Patients: Results from the Phase IV MITO16A/ManGO OV-2 Translational Study.

Simple SummaryThe possibility to identify, with appropriate biomarkers, patients that might mostly benefit from any given treatment is the basis of personalized oncology. Cancer biomarkers should be properly identified and validated on a large number of patients possibly enrolled in dedicated clinical trials. Here, we report the first molecular results of the MITO16A-ManGo-OV2 phase IV trial that was specifically designed to identify prognostic biomarkers of survival in epithelial ovarian cancer patients treated in first line with carboplatin-paclitaxel plus Bevacizumab (NCT01706120), a treatment for which validated predictive or prognostic biomarkers are still lacking. With this work we propose not only novel possible biomarkers for Bevacizumab-treated patients but also a way through which they can be properly collected, analyzed and statistically evaluated in the frame of large multicenter clinical trials.Background. Epithelial ovarian cancer (EOC) is a rare, highly lethal disease. In a subset of high grade EOC patients, maintenance therapy with the antiangiogenic drug Bevacizumab (BEV) is a valuable option. To date, no validated predictive or prognostic biomarkers exist for selecting EOC patients that might benefit from BEV treatment. Methods. Immunohistochemistry and RT-qPCR evaluated the expression of seven angiogenesis-related proteins and of a twelve microRNAs angio-signature in EOC patients, treated in first line with chemotherapy plus BEV (MITO16A/ManGO OV-2 phase IV trial). Centralized statistical analyses assessed the associations between each biomarker, clinical prognostic factors and survival outcomes. Results. High miR-484 expression was associated with longer progression-free and overall survival. Notably, the combined expression of miR-484 and its target VEGFB identified a subset of patients that might mostly benefit from BEV treatment. No other significant correlations were found between the other analyzed biomarkers and patients’ survival. The application of a shrinkage procedure to adjust for over-fitting hazard ratio estimates reduced the association significance. Conclusions. The analysis of angiogenesis related biomarkers in EOC patients homogenously treated with BEV in first line provides novel insight in their prognostic value and suggests that some of them might merit to be tested as predictive markers of drug activity in dedicated randomized trials.

International Vulnerability of Sub-Saharan African Countries with Fixed Exchange Regimes

The fast-growing of the sub-Saharan Africa (SSA) countries over the past decade highlight that successful economic integration can assume many forms, instead of standard patterns. In this sense, this paper provides empirical evidence of the vulnerability of the West African Economic and Monetary Union (WAEMU) and Cape Verde, that adopted fixed exchange rate regimes. For this reason, we developed a global vector autoregressive GVAR model, a data shrinkage procedure, where large sets of variables are reduced to a small set of factors. It allows describing systems of higher dimensional complexities and simulating shocks in a cross-regional approach with the use of common factors. First, we specified and identified a weight matrix for the selected countries, used to reproduce this policy framework, conditioning the parameters to an interdependent trading structure. Next, a variety of tests were estimated, e.g., stationarity, structural break tests and exogeneity tests. We use the generalized impulse response functions to simulate shocks originating from European Monetary Union (EMU), China and the United States (US) on GDP, the CPI, the interest rate and the exchange rate of the SSA countries. Our results were also controlled by the productivity structures of the countries, characterized as agricultural-exporting, commodity-importing, commodity-exporting, and trading based types. In decreasing order of magnitude, shocks originating from Europe (-0.94%), China (-0.73%) and the US (-0.25%) led to a persistent decrease in the GDP and to depreciation of the exchange rate of the region. However, they were not significant for inflation. Shocks in interest rates suggest passive monetary policy. While our empirical evidence does not indicate distinct effects of shocks emanating from the economy-type structures, it points out that these structures are clearly vulnerable to external shocks. To reduce uncertainties and transaction costs from the EMU, SSA countries may consider gradually abandoning the fixed exchange rate regime with the Euro in the medium term and anticipate the development of new multilateral trade agreements. In the long term, it is highly indicated that these countries invest in export diversification and structural transformation.

ESG Risk Rating of Alternative Portfolios

Our aim is to propose a methodology to derive an ESG risk rating for any alternative portfolio, from the ESG ratings of its constituents. To do so, we introduce the notion of ESG risk and define the ESG rating as the ratio between the portfolio’s ESG risk and its total risk. We show how both ESG and total risk can be computed with a variance-based formula. The empirical correlation matrix has to go through a prior shrinkage procedure to ensure that the ESG risk cannot be larger than the total risk. An important consequence of our methodology is that a fund trading an investment universe of poorly (resp. highly) ESG-rated instruments, should have a poor (resp. good) ESG rating. We test this methodology on the Lyxor Epsilon Global Trend Fund and observe that results are consistent with qualitative expectations.

Beamspace Channel Estimation for Massive MIMO mmWave Systems: Algorithm and VLSI Design

Millimeter-wave (mmWave) communication in combination with massive multiuser multiple-input multiple-output (MU-MIMO) enables high-bandwidth data transmission to multiple users in the same time-frequency resource. The strong path loss of wave propagation at such high frequencies necessitates accurate channel state information to ensure reliable data transmission. We propose a novel channel estimation algorithm called BEAmspace CHannel EStimation (BEACHES), which leverages the fact that wave propagation at mmWave frequencies is predominantly directional. BEACHES adaptively denoises the channel vectors in the beamspace domain using an adaptive shrinkage procedure that relies on Stein's unbiased risk estimator (SURE). Simulation results for line-of-sight (LoS) and non-LoS mmWave channels reveal that BEACHES performs on par with state-of-the-art channel estimation methods while requiring orders-of-magnitude lower complexity. To demonstrate the effectiveness of BEACHES in practice, we develop a very large-scale integration (VLSI) architecture and provide field-programmable gate array (FPGA) implementation results. Our results show that adaptive channel denoising can be performed at high throughput and in a hardware-friendly manner for massive MU-MIMO mmWave systems with hundreds of antennas.

Predictor Selection Using Lasso to Examine the Association of Motor Proficiency, Postural Control, Visual Efficiency, and Behavior With the Academic Skills of Elementary School-Aged Children

The associations between motor proficiency, postural control, and visual efficiency with teacher and parent ratings of children’s behavior and academic skills were examined among a sample of elementary school-aged children (N = 50). Teacher and parent ratings of students’ academic skills were analyzed separately to examine the predictive nature of measures of motor proficiency. Spearman rank order correlations reported low to moderate relationships between motor proficiency, postural control, and children’s behavior (e.g., ADHD inattention) and academic skills. The least absolute shrinkage and selection operator (lasso) procedure is demonstrated as an approach for variable selection of measures of children’s motor proficiency, postural control, and visual efficiency to predict academic skills. For teacher and parent ratings of academic skills, ADHD symptom of inattention was the strongest model predictor, whereas directional control (postural control) was also a predictor for parent ratings. Study findings shed light on practical and methodological factors associated with motor skills in educational contexts.

Endocrine disrupting chemicals mixture and pregnancy glucose levels among women from a fertility clinic

TPS 743: Health effects in pregnancy, Exhibition Hall, Ground floor, August 26, 2019, 3:00 PM - 4:30 PM Background/Aim: Endocrine disrupting chemicals (EDCs) have been linked to pregnancy complications such as gestational diabetes (GDM). As individuals are exposed to multiple chemicals simultaneously, investigating EDCs as a mixture is required to provide a more realistic assessment. We evaluated the association between pregnancy exposure to EDC mixture and glucose levels in the late second trimester of pregnancy, an established risk factor of GDM. Methods: We used data from 295 women in the Environment and Reproductive Health (EARTH) Study who had data available on 1st and/or 2nd trimester urinary concentrations of 19 EDCs, as well as late 2nd trimester glucose levels. Several methods were used to evaluate chemical mixtures including: weighted quantile sum (WQS) regression, for estimating exposures’ weights in contributing to the mixture-outcome association; shrinkage procedures (LASSO/elastic net) and Bayesian Kernel Machine Regression (BKMR), for variable selection. The 19 chemicals (3 parabens, triclosan, 11 phthalate metabolites, bisphenol A, benzophenone-3, 2,4-dichlorophenol, 2,5-dichlorophenol) concentrations were adjusted for specific gravity and log transformed. All analyses were adjusted for race, education, BMI, age, physical activity, smoking, and infertility diagnosis. A final linear regression model mutually including the selected exposures was further estimated. Results: Focusing on early 2nd trimester exposures, the following chemicals were identified as potential predictors of higher glucose by at least 2 statistical approaches: bisphenol-A, 2,4-dichlorophenol (2,4D), monoethyl phthalate (MEP), and the molar sum of four di(2-ethylhexyl) phthalate metabolites (∑DEHP). When fitting a final model that included these chemicals, suggestive positive associations were observed (e.g., β=2.2 mg/dL for (a unit increase in) log(2,4D), β =5.4 mg/dL for log(MEP), β =3.5 mg/dL for log(∑DEHP)). Analyses on 1st trimester exposures showed similar results. Conclusion: By evaluating EDCs as an environmental mixture, we identified relevant potential risk factors providing important insights in understanding the relationship between this class of chemicals and pregnancy glucose levels.

Bayesian adaptive lasso for additive hazard regression with current status data.

Variable selection is a crucial issue in model building and it has received considerable attention in the literature of survival analysis. However, available approaches in this direction have mainly focused on time-to-event data with right censoring. Moreover, a majority of existing variable selection procedures for survival models are developed in a frequentist framework. In this article, we consider additive hazards model in the presence of current status data. We propose a Bayesian adaptive least absolute shrinkage and selection operator procedure to conduct a simultaneous variable selection and parameter estimation. Efficient Markov chain Monte Carlo methods are developed to implement posterior sampling and inference. The empirical performance of the proposed method is demonstrated by simulation studies. An application to a study on the risk factors of heart failure disease for type 2 diabetes patients is presented.

Regularized semiparametric estimation of high dimensional dynamic conditional covariance matrices

A three-step estimation strategy for dynamic conditional correlation (DCC) models is proposed. In the first step, conditional variances for individual and aggregate series are estimated by means of QML equation by equation. In the second step, conditional covariances are estimated by means of the polarization identity and conditional correlations are estimated by their usual normalization. In the third step, the two-step conditional covariance and correlation matrices are regularized by means of a new non-linear shrinkage procedure and optimally smoothed. Due to its scant computational burden, the proposed regularized semiparametric DCC model (RSP-DCC) allows to estimate high dimensional conditional covariance and correlation matrices. An application to global minimum variance portfolio is also provided, confirming that RSP-DCC is a simple and viable alternative to existing DCC models.

Regularized Semiparametric Estimation of High Dimensional Dynamic Conditional Covariance Matrices

This paper proposes a three-step estimation strategy for dynamic conditional correlation models. In the first step, conditional variances for individual and aggregate series are estimated by means of QML equation by equation. In the second step, conditional covariances are estimated by means of the polarization identity, and conditional correlations are estimated by their usual normalization. In the third step, the two-step conditional covariance and correlation matrices are regularized by means of a new non-linear shrinkage procedure and used as starting value for the maximization of the joint likelihood of the model. This yields the final, third step smoothed estimate of the conditional covariance and correlation matrices. Due to its scant computational burden, the proposed strategy allows to estimate high dimensional conditional covariance and correlation matrices. An application to global minimum variance portfolio is also provided, confirming that SP-DCC is a simple and viable alternative to existing DCC models.

Detecting Bubbles in the US and UK Real Estate Markets

This study considers state of the art subset selection and shrinkage procedures − stepwise regression, ridge regression, lasso, bridge regression and the elastic net along with the commonly employed least squares regression − to detect bubbles in real estate markets. Our analysis of real estate indices representing the commercial, residential and equity real estate sectors in the United States and the United Kingdom finds evidence suggesting the existence of significant periods of overvaluation in residential real estate, as well as economically significant periods of undervaluation in equity real estate markets. The evolution of specific real estate indices in the United States is similar to the evolution of the corresponding indices in the United Kingdom. In order to determine whether the observed deviations of the actual price index from its fundamental value are due to the presence of bubbles, we use two complementary methodologies, the first based on right-side unit root tests for explosive behaviour and the second defined by regime switching models for bubbles. We show that employing an average of all complex models yields more robust forecasting over an 8 years out-of-sample period.

Improved Shrinkage Estimators of Covariance Matrices With Toeplitz-Structured Targets in Small Sample Scenarios

Shrinkage regularization is an effective strategy to estimate the covariance matrix of multi-variate random vector in small sample scenarios. The purpose of this paper is to propose improved linear shrinkage estimators of covariance matrix as two types of Toeplitz-structured target matrices are respectively employed in the shrinkage procedure. Under Gaussian and non-Gaussian distributions, the corresponding shrinkage estimators are respectively obtained in closed form by unbiasedly estimating the unknown scalar quantities which involve the true covariance matrix. Compared with the existing estimators of same type, the proposed covariance estimators show a significant improvement on the mean squared error in numerical simulations. Moreover, example applications including portfolio risk estimation and classification of real data are provided for verifying the performance of proposed covariance estimators in small sample scenarios.

Suppressing Numerical Oscillation for Nonlinear Hyperbolic Equations by Wavelet Analysis

In the numerical solution for nonlinear hyperbolic equations, numerical oscillation often shows and hides the real solution with the progress of computation. Using wavelet analysis, a dual wavelet shrinkage procedure is proposed, which allows one to extract the real solution hidden in the numerical solution with oscillation. The dual wavelet shrinkage procedure is introduced after applying the local differential quadrature method, which is a straightforward technique to calculate the spatial derivatives. Results free from numerical oscillation can be obtained, which can not only capture the position of shock and rarefaction waves, but also keep the sharp gradient structure within the shock wave. Three model problems—a one-dimensional dam-break flow governed by shallow water equations, and the propagation of a one-dimensional and a two-dimensional shock wave controlled by the Euler equations—are used to confirm the validity of the proposed procedure.

Simultaneous estimation of deterministic and fractal stochastic components in non-stationary time series

In the past few decades, it has been recognized that 1∕f fluctuations are ubiquitous in nature. The most widely used mathematical models to capture the long-term memory properties of 1∕f fluctuations have been stochastic fractal models. However, physical systems do not usually consist of just stochastic fractal dynamics, but they often also show some degree of deterministic behavior. The present paper proposes a model based on fractal stochastic and deterministic components that can provide a valuable basis for the study of complex systems with long-term correlations. The fractal stochastic component is assumed to be a fractional Brownian motion process and the deterministic component is assumed to be a band-limited signal. We also provide a method that, under the assumptions of this model, is able to characterize the fractal stochastic component and to provide an estimate of the deterministic components present in a given time series. The method is based on a Bayesian wavelet shrinkage procedure that exploits the self-similar properties of the fractal processes in the wavelet domain. This method has been validated over simulated signals and over real signals with economical and biological origin. Real examples illustrate how our model may be useful for exploring the deterministic–stochastic duality of complex systems, and uncovering interesting patterns present in time series.

Structural Variability from Noisy Tomographic Projections.

In cryo-electron microscopy, the three-dimensional (3D) electric potentials of an ensemble of molecules are projected along arbitrary viewing directions to yield noisy two-dimensional images. The volume maps representing these potentials typically exhibit a great deal of structural variability, which is described by their 3D covariance matrix. Typically, this covariance matrix is approximately low rank and can be used to cluster the volumes or estimate the intrinsic geometry of the conformation space. We formulate the estimation of this covariance matrix as a linear inverse problem, yielding a consistent least-squares estimator. For n images of size N-by-N pixels, we propose an algorithm for calculating this covariance estimator with computational complexity , where the condition number κ is empirically in the range 10-200. Its efficiency relies on the observation that the normal equations are equivalent to a deconvolution problem in six dimensions. This is then solved by the conjugate gradient method with an appropriate circulant preconditioner. The result is the first computationally efficient algorithm for consistent estimation of the 3D covariance from noisy projections. It also compares favorably in runtime with respect to previously proposed nonconsistent estimators. Motivated by the recent success of eigenvalue shrinkage procedures for high-dimensional covariance matrix estimation, we incorporate a shrinkage procedure that improves accuracy at lower signal-to-noise ratios. We evaluate our methods on simulated datasets and achieve classification results comparable to state-of-the-art methods in shorter running time. We also present results on clustering volumes in an experimental dataset, illustrating the power of the proposed algorithm for practical determination of structural variability.