Conventional Parametric Models Research Articles

Extrapolation of Survival Data Using a Bayesian Approach: A Case Study Leveraging External Data from Cilta-Cel Therapy in Multiple Myeloma.

Extrapolating long-term overall survival (OS) from shorter-term clinical trial data is key to health technology assessment in oncology. However, extrapolation using conventional methods is often subject to uncertainty. Using ciltacabtagene autoleucel (cilta-cel), a chimeric antigen receptor T-cell therapy for multiple myeloma, we used a flexible Bayesian approach to demonstrate use of external longer-term data to reduce the uncertainty in long-term extrapolation. The pivotal CARTITUDE-1 trial (NCT03548207) provided the primary efficacy data for cilta-cel, including a 12-month median follow-up snapshot of OS. Longer-term (48-month median follow-up) survival data from the phase I LEGEND-2 study (NCT03090659) were also available. Twelve-month CARTITUDE-1 OS data were extrapolated in two ways: (1) conventional survival models with standard parametric distributions (uninformed), and (2) Bayesian survival models whose shape prior was informed from 48-month LEGEND-2 data. For validation, extrapolations from 12-month CARTITUDE-1 data were compared with observed 28-month CARTITUDE-1 data. Extrapolations of the 12-month CARTITUDE-1 data using conventional uninformed parametric models were highly variable. Using informative priors from the 48-month LEGEND-2 dataset, the ranges of projected OS at different timepoints were consistently narrower. Area differences between the extrapolation curves and the 28-month CARTITUDE-1 data were generally lower in informed Bayesian models, except for the uninformed log-normal model, which had the lowest difference. Informed Bayesian survival models reduced variation of long-term projections and provided similar projections as the uninformed log-normal model. Bayesian models generated a narrower and more plausible range of OS projections from 12-month data that aligned with observed 28-month data. CARTITUDE-1 ClinicalTrials.gov identifier, NCT03548207. LEGEND-2 ClinicalTrials.gov identifier, NCT03090659, registered retrospectively on 27 March 2017, and ChiCTR-ONH-17012285.

Abnormal structural and functional network topological properties associated with left prefrontal, parietal, and occipital cortices significantly predict childhood TBI-related attention deficits: A semi-supervised deep learning study.

Traumatic brain injury (TBI) is a major public health concern in children. Children with TBI have elevated risk in developing attention deficits. Existing studies have found that structural and functional alterations in multiple brain regions were linked to TBI-related attention deficits in children. Most of these existing studies have utilized conventional parametric models for group comparisons, which have limited capacity in dealing with large-scale and high dimensional neuroimaging measures that have unknown nonlinear relationships. Nevertheless, none of these existing findings have been successfully implemented to clinical practice for guiding diagnoses and interventions of TBI-related attention problems. Machine learning techniques, especially deep learning techniques, are able to handle the multi-dimensional and nonlinear information to generate more robust predictions. Therefore, the current research proposed to construct a deep learning model, semi-supervised autoencoder, to investigate the topological alterations in both structural and functional brain networks in children with TBI and their predictive power for post-TBI attention deficits. Functional magnetic resonance imaging data during sustained attention processing task and diffusion tensor imaging data from 110 subjects (55 children with TBI and 55 group-matched controls) were used to construct the functional and structural brain networks, respectively. A total of 60 topological properties were selected as brain features for building the model. The model was able to differentiate children with TBI and controls with an average accuracy of 82.86%. Functional and structural nodal topological properties associated with left frontal, inferior temporal, postcentral, and medial occipitotemporal regions served as the most important brain features for accurate classification of the two subject groups. Post hoc regression-based machine learning analyses in the whole study sample showed that among these most important neuroimaging features, those associated with left postcentral area, superior frontal region, and medial occipitotemporal regions had significant value for predicting the elevated inattentive and hyperactive/impulsive symptoms. Findings of this study suggested that deep learning techniques may have the potential to help identifying robust neurobiological markers for post-TBI attention deficits; and the left superior frontal, postcentral, and medial occipitotemporal regions may serve as reliable targets for diagnosis and interventions of TBI-related attention problems in children.

Parametric modelling of nearshore wave reflection

The importance of wave reflection on nearshore hydrodynamics has recently been acknowledged by an increasing number of studies, yet it is difficult to simulate by the conventional parametric wave transformation models widely used in coastal engineering practice. Here, we propose a simple and efficient semi-empirical method to deal with this issue. The proposed method considers wave reflections resulted from both the underwater sloping bottom and the shoreline. The former is represented by a local underwater reflection-induced energy transfer term, based on a similarity analysis of the bulk wave reflection from equivalent sloping steps. The latter is calculated as the shoreline boundary condition by utilizing a simple empirical formula of wave reflection coefficient for an emerged coastal structure. Model verifications against the previous theoretical/empirical formulas, experimental data and numerical results suggest a promising predictability of the proposed method for the cross-shore reflected wave height transformation with various bathymetry configurations. Three different formulas of the bulk reflection coefficient for an underwater sloping step are tested and compared, indicating the importance of considering slope effects in these formulas. Model results also suggest that wave reflection along a steep beach profile can be quantitatively comparable to the shoreline reflection, and it is important to include reflections of both sources in numerical modelling. The simple method can be extended and incorporated into the existing operational parametric wave models. It serves as a possible approach to simulate wave reflection with low computational demand and reasonable accuracy, and to include wave reflection influences in studying coastal processes on a broad range of time and spatial scales.

Leveraging machine learning methods to estimate heterogeneous effects: father absence in China as an example

Individuals differ in their personal and environmental characteristics, and the same treatment or condition may affect individuals in different ways or magnitudes. Heterogeneity in effects thus has important implications for academic research and policymaking. However, it is difficult to uncover and estimate heterogenous effects using conventional parametric models without making assumptions based on limited information, and results can be difficult to interpret when involving a large number of moderators. To address these limitations, this paper introduces three supervised machine learning methods for estimating heterogeneous treatment effects with experimental and observational data: causal forest, Bayesian Additive Regression Trees (BART), and an ensemble approach called X-learner. These methods are first applied to simulated datasets and then implemented using empirical education survey data from China to estimate heterogeneous effects of father absence on student cognitive ability across a series of individual and family characteristics.

Deep generative models for galaxy image simulations

ABSTRACT Image simulations are essential tools for preparing and validating the analysis of current and future wide-field optical surveys. However, the galaxy models used as the basis for these simulations are typically limited to simple parametric light profiles, or use a fairly limited amount of available space-based data. In this work, we propose a methodology based on deep generative models to create complex models of galaxy morphologies that may meet the image simulation needs of upcoming surveys. We address the technical challenges associated with learning this morphology model from noisy and point spread function (PSF)-convolved images by building a hybrid Deep Learning/physical Bayesian hierarchical model for observed images, explicitly accounting for the PSF and noise properties. The generative model is further made conditional on physical galaxy parameters, to allow for sampling new light profiles from specific galaxy populations. We demonstrate our ability to train and sample from such a model on galaxy postage stamps from the HST/ACS COSMOS survey, and validate the quality of the model using a range of second- and higher order morphology statistics. Using this set of statistics, we demonstrate significantly more realistic morphologies using these deep generative models compared to conventional parametric models. To help make these generative models practical tools for the community, we introduce galsim-hub, a community-driven repository of generative models, and a framework for incorporating generative models within the galsim image simulation software.

Do Machine Learning Methods Outperform Traditional Statistical Models in Crime Prediction? A Comparison Between Logistic Regression and Neural Networks

Although machine learning (ML) methods have recently gained popularity in both academia and industry as alternative risk assessment tools for efficient decision-making, inconsistent patterns are observed in the existing literature regarding their competitiveness and utility in predicting various outcomes. Drawing on a sample of the general youth population in the U.S., we compared the predictive accuracy of logistic regression (LR) and neural networks (NNs), which are the most widely applied approaches in conventional statistics and contemporary ML methods, respectively, by adopting many theoretically relevant predictors of the future arrest outcome. Even after fully implementing rigorous ML protocols for model tuning and up-sampling and down-sampling procedures recommended in recent literature to optimize learning algorithms, NNs did not yield substantially improved performance over LR if we still rely on a conventional dataset with relatively small sample sizes and a limited number of predictors. Nonetheless, we encourage more rigorous, comprehensive, and diverse evaluation research for a complete understanding of the ML potential in predictive capacity and the contingencies in which modern ML methods can perform better than conventional parametric statistical models.

Modeling Complex Pharmacokinetics of Long‐Acting Injectable Products Using Convolution‐Based Models With Nonparametric Input Functions

The interest in the development and the therapeutic use of long‐acting injectable (LAI) products for chronic or long‐term treatments has grown exponentially. The complexity and the multiphase drug release process represent serious issues for an effective modeling of the PK properties of LAI products. The objective of this article is to show how convolution‐based models with piecewise‐linear approximation of the nonlinear drug release function can provide an enhanced modeling tool for (1) characterizing the complex PK profiles of LAI formulations with completely different drug release properties, and (2) addressing key questions supporting the optimal development of LAI products by simulating the PK time course resulting from different dosing strategies. Convolution‐based modeling and simulation were implemented in NONMEM, and 3 case studies were presented to assess the performances of this new modeling approach using PK data of LAI products developed using different technologies and administered using different routes: microsphere technology and aqueous nanosuspension intramuscularly administered and biodegradable polymer subcutaneously administered. The performance of the convolution‐based modeling approach was compared with the performance of conventional parametric models using a reference data set on theophylline. The results of the comparison indicated that the nonparametric input function provided a more accurate description of the data either in terms of global measure of goodness of fit (ie, Akaike information criterion and Bayesian information criterion) or in terms of performance of the fitted model (ie, the percent prediction error on Cma x and AUC0‐t).

Background Harmonic Probabilistic Model for Harmonic Responsibility Assessment Based on Nonparametric Methods

As the increasing integration of inverter‐interfaced renewable energy generators intensifies the volatility and uncertainty of background harmonics in the utility system, it is crucial to construct a precise background harmonic model and determine the harmonic responsibility to promote harmonic management. This paper proposes a probabilistic model of background harmonics for harmonic responsibility assessment based on nonparametric methods. First, the amplitude data of background harmonic voltage are segmented in chronological order considering the stochastic features. Then, the nonparametric Bootstrap resampling is introduced to handle the case of insufficient samples, and the Kernel density estimation with optimal bandwidth is used to depict the probability distribution characteristics in each time interval. For the sake of identifying the piecewise probability density function with optimal fitting effect, the goodness‐of‐fit test is subsequently applied. Finally, based on the proposed background harmonic probabilistic model, the harmonic responsibility assessment is conducted without the involvement of phase angle information. The detailed performance verification of the proposed model is carried out in light of the field‐measured data. The results validate that the proposed model outperforms the conventional parametric models in handling insufficient data and fitting accuracy and is helpful to achieve the effective quantification of the overall harmonic responsibility within a period of time. © 2020 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.

Identification of nonlinear kinetics of macroscopic bio-reactions using multilinear Gaussian processes

In biological systems, nonlinear kinetic relationships between metabolites of interest are modeled for various purposes. Usually, little a priori knowledge is available in such models. Identifying the unknown kinetics is, therefore, a critical step which can be very challenging due to the problems of (i) model selection and (ii) nonlinear parameter estimation. In this paper, we aim to address these problems systematically in a framework based on multilinear Gaussian processes using a family of kernels tailored to typical behaviours of modulation effects such as activation and inhibition or combinations thereof. Using one such process as a model for each modulation effect leads to a much more flexible model than conventional parametric models, e.g., the Monod model. The resulting models of the modulation effects can also be used as a starting point for estimating parametric kinetic models. As each modulation effect is modeled separately, this task is greatly simplified compared to the conventional approach where the parameters in all modulation functions have to be estimated simultaneously. We also show how the type of modulation effect can be selected automatically by way of regularization, thus by-passing the model selection problem. The resulting parameter estimates can be used as initial estimates in the conventional approach where the full model is estimated. Numerical experiments, including fed-batch simulations, are conducted to demonstrate our methods.

Is there an EKC between economic growth and smog pollution in China? New evidence from semiparametric spatial autoregressive models

Although China has achieved rapid economic growth, it suffers from severe smog conditions. Determining how to promote cleaner production and environmental sustainability is of great significance in clarifying the relationship between income and pollutants. Within an extension of the stochastic impacts by regression on population, affluence and technology (STIRPAT) framework, the effects of economic growth on PM2.5 contamination are explicitly investigated based on a recent large sample of 249 Chinese cities in 2015. From a nonlinear spatial perspective, this study marks the first attempt at comprehensively exploring the nexus between the selected variables and PM2.5 emissions using semiparametric spatial autoregressive models, which overcome the misspecification issues of conventional parametric models. To avoid the limitations of a single method, this study also applies different estimation techniques such as the spatial lag model (SLM), the spatial autoregressive model with spatial autoregressive disturbances (SARAR), two-stage least squares regression (2SLSR), quantile regression (QR) and nonparametric regression (NPR). The results reveal that PM2.5 pollutants show strong positive characteristics of spatial spillover. The results also suggest that there is a significant inverted U-shaped relationship between economic growth and PM2.5 concentrations, which confirms the environmental Kuznets curve (EKC) hypothesis. Additionally, the empirical findings highlight the influence that the population density, industrialization, urbanization and traffic development factors have on increasing PM2.5 emissions. In contrast, technological innovation and green coverage do not play important roles in reducing PM2.5 concentrations. The distribution of PM2.5 concentrations displays apparent geographical features. To provide an in-depth understanding of these impacts, marginal analysis is performed to distinguish the local and neighbouring effects of determinants on PM2.5 pollution.

Robust determination of differential abundance in shotgun proteomics using nonparametric statistics.

Label-free shotgun mass spectrometry enables the detection of significant changes in protein abundance between different conditions. Due to often limited cohort sizes or replication, large ratios of potential protein markers to number of samples, as well as multiple null measurements pose important technical challenges to conventional parametric models. From a statistical perspective, a scenario similar to that of unlabeled proteomics is encountered in genomics when looking for differentially expressed genes. Still, the difficulty of detecting a large fraction of the true positives without a high false discovery rate is arguably greater in proteomics due to even smaller sample sizes and peptide-to-peptide variability in detectability. These constraints argue for nonparametric (or distribution-free) tests on normalized peptide values, thus minimizing the number of free parameters, as well as for measuring significance with permutation testing. We propose such a procedure with a class-based statistic, no parametric assumptions, and no parameters to select other than a nominal false discovery rate. Our method was tested on a new dataset which is available via ProteomeXchange with identifier PXD006447. The dataset was prepared using a standard proteolytic digest of a human protein mixture at 1.5-fold to 3-fold protein concentration changes and diluted into a constant background of yeast proteins. We demonstrate its superiority relative to other approaches in terms of the realized sensitivity and realized false discovery rates determined by ground truth, and recommend it for detecting differentially abundant proteins from MS data.

Hedonic Analysis of Housing Sales Prices with Semiparametric Methods

This study estimates a hedonic price function using a semiparametric regression based on Penalized Spline Smoothing, and compares the price prediction performance with conventional parametric models. The excellent results obtained show that the semiparametric models allow to obtain a significant improvement in the prediction of housing sales prices.

Development of a predictive model for study of skin-core phenomenon in stabilization process of PAN precursor

Studying the presence and progress of fiber defects, such as skin-core structure, is an important tool for analysis of a chemical process. In this article, the skin core morphology has been analyzed by optical microscopic (OM) images and Fourier transform infrared attenuated total reflectance mapping (FTIR-ATR mapping). The results of FTIR-ATR mapping showed that the fiber is almost uniform in the core area while OM images are accurate enough to be used for skin-core analysis. Using OM images, the core ratio of samples were measured to quantify the skin-core structure. Non-parametric kernel density estimation methods have then been compared with conventional parametric distribution models using these data. The results reveal that the parametric methods cannot adequately describe the skin-core phenomenon and that the non-parametric distributions are more appropriate for the quantification of skin-core morphology. By applying the non-parametric distributions, a model has been developed, which describes the relationship between the skin-core defect and the operation parameters of the fiber production. This approach can be used to predict the probability of skin-core occurrence and can be used to decrease the presence of this phenomenon in the carbon fibers production industry. Our results show that temperature is one of the most significant operational parameter at a typical oxygen concentration (in air at atmospheric pressure) governing the skin-core formation.

Integrated Framework to Study Efficient Spectral Estimation Techniques for Assessing Spectral Efficiency Analysis

The advanced network applications enable software driven spectral analysis of non-stationary signal or processes which precisely involves domain analysis with the purpose of decomposing a complex signal coefficients into simpler forms. However, the proper estimation of power coefficients over frequency components of a random signal leads to provide very useful information required in various fields of study. The complex design constraints associated with conventional parametric models such as Dynamic Average Model, Autoregressive MA, etc. for multidimensional spectral estimation using adaptive filters leads to a situation where higher computational complexities generate significant overhead on the systems. Therefore, the proposed study aims to formulate an efficient framework intended to derive a fast algorithm for processing Adaptive Capon and Phase Estimator (APES). The proposed method is applied to a non-stationary signal which is random. Further, the adaptive estimation of power spectra along with more accurate spectral efficiency has been identified in case of APES. An extensive performance evaluation followed by a comparative analysis has been performed by obtaining the values from different spectral estimation techniques, such as APES, PSC, ASC, and CAPON. Moreover, the framework ensures that unlike others, APES is subjected to attain superior signal quality regarding Power Spectral Density (PSD) and Signal to Noise Ratio (SNR) while achieving very less amount of Mean Square Error (MSE). It also exhibits comparatively low convergence speed and computational complexity as compared to its legacy versions.

Nonparametric machine learning models for predicting the credit default swaps: An empirical study

Credit default swap which reflects the credit risk of a firm is one of the most frequently traded credit derivatives. In this paper, we conduct a comprehensive study to verify the predictive performance of nonparametric machine learning models and two conventional parametric models on the daily credit default swap spreads of different maturities and different rating groups, from AA to C. The whole period of data set used in this study runs from January 2001 to February 2014, which includes the global financial crisis period when the credit risk of firms were very high. Through experiments, it is shown that most nonparametric models used in this study outperformed the parametric benchmark models in terms of prediction accuracy as well as the practical hedging measures irrespective of the different credit ratings of the firms and the different maturities of their spreads. Especially, artificial neural networks showed better performance than the other parametric and nonparametric models.

A CONDITIONALLY PARAMETRIC PROBIT MODEL OF MICRODATA LAND USE IN CHICAGO

ABSTRACTSpatial data sets pose challenges for discrete choice models because the data are unlikely to be independently and identically distributed. A conditionally parametric spatial probit model is amenable to very large data sets while imposing far less structure on the data than conventional parametric models. We illustrate the approach using data on 474,170 individual lots in the City of Chicago. The results suggest that simple functional forms are not appropriate for explaining the spatial variation in residential land use across the entire city.

Semiparametric nonlinear regression for detecting gene and environment interactions

It is commonly accepted that gene and environment (G×E) interactions play a pivotal role in determining the risk of human diseases. In conventional parametric models such as linear models and generalized linear models which are applied frequently to study statistical interactions, effects of covariates are decomposed into main effects and interaction effects (products of two components). Such decomposition, however, may not reflect the true interaction effect of gene and environment. In this paper, we propose a semiparametric regression approach to capture possible nonlinear G×E interactions. A profile quasi-log-likelihood estimation method is applied with asymptotic consistency and normality established for the profile estimators. Moreover, we develop Rao-score-type test procedures based on the profile estimation for regression parameters and nonparametric coefficient functions, respectively. Our models and methods are illustrated by both simulation studies and analysis of a dataset application.

Semiparametric spatial effects kernel minimum squared error model for predicting housing sales prices

Semiparametric regression models have been extensively used to predict housing sales prices, but semiparametric kernel machines with spatial effect have not been studied yet. This paper proposes the semiparametric spatial effect kernel minimum squared error model (SSEKMSEM) and the semiparametric spatial effect least squares support vector machine (SSELS-SVM) for estimating a hedonic price function and compares the price prediction performance with the conventional parametric models and a semiparametric generalized additive model (GAM). This paper utilizes two data sets. One is a large data set representing 5966 single-family residential home sales between July 2000 and August 2008 from Pitt County, North Carolina. The other is a data set of residential property sales records from September 2000 to September 2004 in Carteret County, North Carolina. The results show that the SSEKMSEM and SSELS-SVM outperform the parametric counterparts and the semiparametric GAM in both in-sample and out-of-sample price predictions, indicating that these kernel machines can be useful for measurement and prediction of housing sales prices.

Statistical Procedures for Forecasting Criminal Behavior

There is a substantial and powerful literature in statistics and computer science clearly demonstrating that modern machine learning procedures can forecast more accurately than conventional parametric statistical models such as logistic regression. Yet, several recent studies have claimed that for criminal justice applications, forecasting accuracy is about the same. In this paper, we address the apparent contradiction. Forecasting accuracy will depend on the complexity of the decision boundary. When that boundary is simple, most forecasting tools will have similar accuracy. When that boundary is complex, procedures such as machine learning, that proceed adaptively from the data will improve forecasting accuracy, sometimes dramatically. Machine learning has other benefits as well, and e↵ective software is readily available.

Overview of: “Statistical Procedures for Forecasting Criminal Behavior: A Comparative Assessment”

A substantial and powerful literature in statistics and computer science has clearly demonstrated that modern machine learning procedures can forecast more accurately than conventional parametric statistical models such as logistic regression. Yet, several recent studies have claimed that for criminal justice applications, forecasting accuracy is about the same. In this article, we address the apparent contradiction. Forecasting accuracy will depend on the complexity of the decision boundary. When that boundary is simple, most forecasting tools will have similar accuracy. When that boundary is complex, procedures such as machine learning, which proceed adaptively from the data, will improve forecasting accuracy, sometimes dramatically. Machine learning has other benefits as well, and effective software is readily available. The complexity of the decision boundary will in practice be unknown, and there can be substantial risks to gambling on simplicity. Criminal justice decision makers and other stakeholders can be seriously misled with rippling effects going well beyond the immediate offender. There seems to be no reason for continuing to rely on traditional forecasting tools such as logistic regression.