When it is suspected that the treatment effect may only be strong for certain subpopulations, identifying the baseline covariate profiles of subgroups who benefit from such a treatment is of key importance. In this pa...
详细信息
When it is suspected that the treatment effect may only be strong for certain subpopulations, identifying the baseline covariate profiles of subgroups who benefit from such a treatment is of key importance. In this paper, we propose an approach for subgroup analysis by firstly introducing Bernoulli-gated hierarchical mixtures of experts (BHME), a binary-tree structured model to explore heterogeneity of the underlying distribution. We show identifiability of the BHME model and develop an em-based maximum likelihood method for optimization. The algorithm automatically determines a partition structure with optimal prediction but possibly suboptimal in identifying treatment effect heterogeneity. We then suggest a testing-based postscreening step to further capture effect heterogeneity. Simulation results show that our approach outperforms competing methods on discovery of differential treatment effects and other related metrics. We finally apply the proposed approach to a real dataset from the Tennessee's Student/Teacher Achievement Ratio project.
Nonparametric estimation methods for the cure rate and the distribution of the failure time of uncured subjects with covariates for censored survival data have attracted much attention in the last few years. To model ...
详细信息
Nonparametric estimation methods for the cure rate and the distribution of the failure time of uncured subjects with covariates for censored survival data have attracted much attention in the last few years. To model the effects of covariates on the distribution of the failure time of uncured subjects, existing works assume that the cure rate is a constant or depends on the same covariate as the distribution of uncured subjects. In this paper, we review the nonparametric estimation methods in the context of the mixture cure model and propose a new nonparametric estimator for the distribution of uncured subjects that relaxes the assumption used in the existing works. The estimation is based on the em algorithm, which is readily available for mixture cure models, and is strongly consistent. The finite sample performance of the proposed estimator is assessed and compared with existing methods in a simulation study. Finally, the nonparametric estimation methods are employed to model the effects of some covariates on the time to bankruptcy among commercial banks insured by the Federal Deposit Insurance Corporation during the first quarter of 2006.
A new survival model is proposed in the presence of surviving fractions and unobserved dispersion. It is obtained by considering several latent factors (or risks) that generated the observed lifetime which follows a g...
详细信息
A new survival model is proposed in the presence of surviving fractions and unobserved dispersion. It is obtained by considering several latent factors (or risks) that generated the observed lifetime which follows a generalized Poisson distribution, and it includes as a special case, the promotion time cure model. We explore maximum likelihood tools for inference issues by aid of the expectation maximization algorithm for estimating the parameters while model discrimination problem is treated by the aid of the likelihood ratio test. The new regression is applied to cervical cancer data to evaluate covariates effects in the cured fraction and non-cured group.
In this article, we present statistical inference of unknown lifetime parameters based on a progressive Type-I interval censored dataset in presence of independent competing risks. A progressive Type-I interval censor...
详细信息
In this article, we present statistical inference of unknown lifetime parameters based on a progressive Type-I interval censored dataset in presence of independent competing risks. A progressive Type-I interval censoring scheme is a generalization of an interval censoring scheme, allowing intermediate withdrawals of test units at the inspection points. We assume that the lifetime distribution corresponding to a failure mode belongs to a log-location-scale family of distributions. Subsequently, we present the maximum likelihood analysis for unknown model parameters. We observe that the numerical computation of the maximum likelihood estimates can be significantly eased by developing an expectation-maximization algorithm. We demonstrate the same for three popular choices of the log-location-scale family of distributions. We then provide Bayesian inference of the unknown lifetime parameters via Gibbs Sampling and a related data augmentation scheme. We compare the performance of the maximum likelihood estimators and Bayesian estimators using a detailed simulation study. We also illustrate the developed methods using a progressive Type-I interval censored dataset.
作者：
Xu, MinQin, ZhongfengBeijing Univ Technol
Fac Sci Sch Stat & Data Sci Beijing Peoples R China Beihang Univ
Sch Econ & Management Beijing Peoples R China Beihang Univ
Key Lab Complex Syst Anal Management & Decis Minist Educ Beijing Peoples R China East China Normal Univ
Key Lab Adv Theory & Applicat Stat & Data Sci Minist Educ Shanghai Peoples R China
Interval-valued data, as typical symbolic data, provide a feasible way to deal with massive data sets. Although a lot of literature has been focused on researching interval-valued regression models, few works are devo...
详细信息
Interval-valued data, as typical symbolic data, provide a feasible way to deal with massive data sets. Although a lot of literature has been focused on researching interval-valued regression models, few works are devoted to exploring Bayesian methods for interval-valued data. In this paper, we propose a novel Bayesian parametrized method for interval-valued data by transforming an interval into a reference point, and further establish a Bayesian linear regression model. The advantage of the Bayesian parametrized method is to make use of full information within intervals and meanwhile it can solve the potential problem of multicollinearity in the parametrized method. We assume the prior distribution is normal with zero mean, and employ the em algorithm to obtain the empirical Bayes estimates. The results of experimental and real data sets show that the Bayesian parametrized method has a superior forecasting advantage when the sample size is small and the random error is large.
Table-based fact verification (TFV) aims to classify whether a statement is entailed or refuted by a given table, which necessitates adept comprehension and logical inference skills among both tabular and textual data...
详细信息
ISBN:
(纸本)9798350359329;9798350359312
Table-based fact verification (TFV) aims to classify whether a statement is entailed or refuted by a given table, which necessitates adept comprehension and logical inference skills among both tabular and textual data. The complexity of TFV is attributed to the involvement of both soft linguistic reasoning and hard symbolic reasoning. Existing studies tend to rely exclusively on table pre-trained models, lacking sufficient reasoning ability and treating various types of reasoning without distinction. In this paper, we propose a novel approach that transforms TFV task into a latent variable learning problem, employing a set of taskspecific functions. Specifically, we leverage the hard ExpectationMaximization (em) algorithm to ascertain the latent logic type underlying statements, then channel each statement through a specialized network designed for unique logical reasoning. Furthermore, we conduct a comprehensive exploration of the practical implementations of our proposed method in TFV task. Our approach diverges from the prevalent reliance on tablebased pre-trained models, yet manages to surpass performance of various baseline models, exemplifying its efficacy and innovation.
Mixture-regularized bidirectional gated recurrent unit with attention (BiGAR) boosts the efficiency of decoding brain signals into hand movement trajectories. The novel neural decoder achieves an R-squared of over 0.8...
详细信息
ISBN:
(纸本)9798350359329;9798350359312
Mixture-regularized bidirectional gated recurrent unit with attention (BiGAR) boosts the efficiency of decoding brain signals into hand movement trajectories. The novel neural decoder achieves an R-squared of over 0.8 in less than 0.2 ms of computation time on the MC_Maze dataset using fewer than 500 training trials. The expectation maximization (em) algorithm used to extract neural hidden states improves R-squared and retains relatively low computation time for BiGAR. We further research on how different mixture regularizers impact the model performance. We generate mixture regularizers through pairwise weighted sum mixing of five individual regularizers associated with the Gaussian, Cauchy, Laplace, Sinc-squared, and Sin-fourth probability density functions. Experiments indicate that the improvement of model R-squared with mixture regularizers exceeds that of traditional individual regularizers and no regularizer.
In this paper, we propose a weighted extension of the family of multivariate Generalized Asymmetric Laplace (GAL) distributions. This extension is formed as a variance-mean mixture of the multivariate normal distribut...
详细信息
In this paper, we propose a weighted extension of the family of multivariate Generalized Asymmetric Laplace (GAL) distributions. This extension is formed as a variance-mean mixture of the multivariate normal distribution and the weighted gamma distribution. By using the weighted gamma distribution as the mixing distribution, the resulting family of weighted GAL (WGAL) distributions gains an additional parameter to further regulate kurtosis and tail thickness;this is an advantage over the family of GAL distributions for modeling data sets. In particular, this new parameter provides great flexibility in adjusting the kurtosis and tail thickness for some members of the GAL distributions family, since these distributions are the widely used members of the GAL family without any shape parameter regulating kurtosis and tail thickness. After defining the multivariate WGAL distributions family and constructing the probability density function, we give some special cases of the new family and examine various properties of the new distributions, such as linear transformations, conditional distributions, and multivariate kurtosis measure. We study the maximum likelihood (ML) estimation to estimate the parameters and describe an algorithm based on the expectation maximization (em) principle to obtain the ML estimates. We also provide simulation studies and real data examples to explore the modeling capacity of some distributions belonging to the newly proposed family of distributions.
This paper introduces the hhsmmR package, which involves functions for initializing, fitting, and predication of hidden hybrid Markov/semi-Markov models. These models are flexible models with both Markovian and semi-M...
详细信息
This paper introduces the hhsmmR package, which involves functions for initializing, fitting, and predication of hidden hybrid Markov/semi-Markov models. These models are flexible models with both Markovian and semi-Markovian states, which are applied to situations where the model involves absorbing or macro-states. The left-to-right models and the models with series/parallel networks of states are two models with Markovian and semi-Markovian states. The hhsmm also includes Markov/semi-Markov switching regression model as well as the auto-regressive HHSMM, the nonparametric estimation of the emission distribution using penalized B-splines, prediction of future states and the residual useful lifetime estimation in the predict function. The commercial modular aero-propulsion system simulation (C-MAPSS) data-set is also included in the package, which is used for illustration of the application of the package features. The application of the hhsmm package to the analysis and prediction of the Spain's energy demand is also presented.
This paper presents a new Bayesian approach to texture classification, yielding enhanced performance in the presence of intraclass diversity. From a mathematical point of view, it specifies an original em algorithm fo...
详细信息
ISBN:
(纸本)9781479983407
This paper presents a new Bayesian approach to texture classification, yielding enhanced performance in the presence of intraclass diversity. From a mathematical point of view, it specifies an original em algorithm for mixture estimation on Riemannian manifolds, generalising existing, non probabilistic, clustering analysis methods. For texture classification, the chosen feature space is the Riemannian manifold known as the Poincare half plane, here denoted H, (this is the set of univariate normal distributions, equipped with Rao's distance). Classes are modelled as finite mixtures of Riemannian priors, (Riemannian priors are probability distributions, recently introduced by the authors, which represent clusters of points in H). During the training phase of classification, the em algorithm, proposed in this paper, computes maximum likelihood estimates of the parameters of these mixtures. The algorithm combines the structure of an em algorithm for mixture estimation, with a Riemannian gradient descent, for computing weighted Riemannian centres of mass.
暂无评论