Overlapping group lasso for high-dimensional generalized linear models

Abstract

Structured sparsity has recently been a very popular technique to deal with the high-dimensional data. In this paper, we mainly focus on the theoretical problems for the overlapping group structure of generalized linear models (GLMs). Although the overlapping group lasso method for GLMs has been widely applied in some applications, the theoretical properties about it are still unknown. Under some general conditions, we presents the oracle inequalities for the estimation and prediction error of overlapping group Lasso method in the generalized linear model setting. Then, we apply these results to the so-called Logistic and Poisson regression models. It is shown that the results of the Lasso and group Lasso procedures for GLMs can be recovered by specifying the group structures in our proposed method. The effect of overlap and the performance of variable selection of our proposed method are both studied by numerical simulations. Finally, we apply our proposed method to two gene expression data sets: the p53 data and the lung cancer data.     

A simulation study of l1:estimation of a seasonal moving average time series model

A detailed simulation study is reported on the application of l₁:estimations to a seasonal moving average model. It is found that the asymptotic normal distribution is a nonapproximation to the finite sample distribution. However, the expected benefits of l₁:estimation relative to l₂:are partially realised for nonnormal innovative distributions.     

Grouping strategies and thresholding for high dimensional linear models

The estimation problem in a high regression model with structured sparsity is investigated. An algorithm using a two-step block thresholding procedure called GR-LOL is provided. Convergence rates are produced: they depend on simple coherence-type indices of the Gram matrix – easily checkable on the data – as well as sparsity assumptions of the model parameters measured by a combination of l₁ within-blocks with _lq$l_q$, q<1$q<1$ between-blocks norms. The simplicity of the coherence indicator suggests ways to optimize the rates of convergence when the group structure is not naturally given by the problem or is unknown. In such a case, an auto-driven procedure is provided to determine the regressor groups (number and contents). An intensive practical study compares our grouping methods with the standard LOL algorithm. We prove that the grouping rarely deteriorates the results but can improve them very significantly. GR-LOL is also compared with group-Lasso procedures and exhibits a very encouraging behavior. The results are quite impressive, especially when GR-LOL algorithm is combined with a grouping pre-processing.     

Estimation of Sobol's sensitivity indices under generalized linear models

We derive explicit formulas for Sobol's sensitivity indices (SSIs) under the generalized linear models (GLMs) with independent or multivariate normal inputs. We argue that the main-effect SSIs provide a powerful tool for variable selection under GLMs with identity links under polynomial regressions. We also show via examples that the SSI-based variable selection results are similar to the ones obtained by the random forest algorithm but without the computational burden of data permutation. Finally, applying our results to the problem of gene network discovery, we identify through the SSI analysis of a public microarray dataset several novel higher-order gene–gene interactions missed out by the more standard inference methods. The relevant functions for SSI analysis derived here under GLMs with identity, log, and logit links are implemented and made available in the R package Sobol sensitivity.     

Another look at linear programming for feature selection via methods of regularization

We consider statistical procedures for feature selection defined by a family of regularization problems with convex piecewise linear loss functions and penalties of l ₁ nature. Many known statistical procedures (e.g. quantile regression and support vector machines with l ₁-norm penalty) are subsumed under this category. Computationally, the regularization problems are linear programming (LP) problems indexed by a single parameter, which are known as ‘parametric cost LP’ or ‘parametric right-hand-side LP’ in the optimization theory. Exploiting the connection with the LP theory, we lay out general algorithms, namely, the simplex algorithm and its variant for generating regularized solution paths for the feature selection problems. The significance of such algorithms is that they allow a complete exploration of the model space along the paths and provide a broad view of persistent features in the data. The implications of the general path-finding algorithms are outlined for several statistical procedures, and they are illustrated with numerical examples.     

Analysis of mixed correlated bivariate zero-inflated count and (k,l)-inflated beta responses with application to social network datasets

This paper presents a new model that monitors the basic network formation mechanisms via the attributes through time. It considers the issue of joint modeling of longitudinal inflated (0, 1)-support continuous and inflated count response variables. For joint model of mentioned response variables, a correlated generalized linear mixed model is studied. The fraction response is inflated in two points k and l (k < l) and a k and l inflated beta distribution is introduced to use as its distribution. Also, the count response is inflated in zero and we use some members of zero-inflated power series distributions, hurdle-at-zero, members of zero-inflated double power series distributions and zero-inflated generalized Poisson distribution as our count response distribution. A full likelihood-based approach is used to yield maximum likelihood estimates of the model parameters and the model is applied to a real social network obtained from an observational study where the rate of the ith node’s responsiveness to the jth node and the number of arrows or edges with some specific characteristics from the ith node to the jth node are the correlated inflated (0, 1)-support continuous and inflated count response variables, respectively. The effect of the sender and receiver positions in an office environment on the responses are investigated simultaneously.     

Bayesian estimation for the M/G/1 queue using a phase-type approximation

This article deals with Bayesian inference and prediction for M/G/1 queueing systems. The general service time density is approximated with a class of Erlang mixtures which are phase-type distributions. Given this phase-type approximation, an explicit evaluation of measures such as the stationary queue size, waiting time and busy period distributions can be obtained. Given arrival and service data, a Bayesian procedure based on reversible jump Markov Chain Monte Carlo methods is proposed to estimate system parameters and predictive distributions.     

Some asymptotic results for hypothesis testing in multivariate failure time data

This paper presents a class of generalized Wald, generalized score and generalized likelihood ratio statistics for hypothesis testing and model selection for multivariate failure time data. These statistics are based on a marginal hazard model with a common baseline hazard function. The large sample distributions of these statistics are examined. It is shown that the proposed test statistics follow asymptotically a weighted sum of independent χ₁² distributions.     

An alternate version of the conceptual predictive statistic based on a symmetrized discrepancy measure

The conceptual predictive statistic, C_p, is a widely used criterion for model selection in linear regression. C_p serves as an estimator of a discrepancy, a measure that reflects the disparity between the generating model and a fitted candidate model. This discrepancy, based on scaled squared error loss, is asymmetric: an alternate measure is obtained by reversing the roles of the two models in the definition of the measure. We propose a variant of the C_p statistic based on estimating a symmetrized version of the discrepancy targeted by C_p. We claim that the resulting criterion provides better protection against overfitting than C_p, since the symmetric discrepancy is more sensitive towards detecting overspecification than its asymmetric counterpart. We illustrate our claim by presenting simulation results. Finally, we demonstrate the practical utility of the new criterion by discussing a modeling application based on data collected in a cardiac rehabilitation program at University of Iowa Hospitals and Clinics.     

A modified Cp statistic in a system-of-equations model

A new statistic, SΓ(p), is developed for variable selection in a system-of-equations model. The standardized total mean square error in the SΓ(p)statistic is weighted by the covariance matrix of dependent variables instead of the error covariance matrix of the true model as in the original definition. The new statistic can be also used for model selection in the non-nested models. The estimate of SΓ(p), SC(p), is derived and shown to become SC_ε(p) in the similar form of C_p in a single-equation model when the covariance matrix of sampled dependent variables is replaced by the error covariance matrix under the full model.     

Robust variable selection for the varying coefficient model based on composite L 1–L 2 regression

The varying coefficient model (VCM) is an important generalization of the linear regression model and many existing estimation procedures for VCM were built on L ₂ loss, which is popular for its mathematical beauty but is not robust to non-normal errors and outliers. In this paper, we address the problem of both robustness and efficiency of estimation and variable selection procedure based on the convex combined loss of L ₁ and L ₂ instead of only quadratic loss for VCM. By using local linear modeling method, the asymptotic normality of estimation is driven and a useful selection method is proposed for the weight of composite L ₁ and L ₂. Then the variable selection procedure is given by combining local kernel smoothing with adaptive group LASSO. With appropriate selection of tuning parameters by Bayesian information criterion (BIC) the theoretical properties of the new procedure, including consistency in variable selection and the oracle property in estimation, are established. The finite sample performance of the new method is investigated through simulation studies and the analysis of body fat data. Numerical studies show that the new method is better than or at least as well as the least square-based method in terms of both robustness and efficiency for variable selection.     

Resistant selection of the smoothing parameter for smoothing splines

Robust automatic selection techniques for the smoothing parameter of a smoothing spline are introduced. They are based on a robust predictive error criterion and can be viewed as robust versions of C _p and cross-validation. They lead to smoothing splines which are stable and reliable in terms of mean squared error over a large spectrum of model distributions.     

Linear model selection by cross-validation

We consider the problem of model (or variable) selection in the classical regression model based on cross-validation with an added penalty term for penalizing overfitting. Under some weak conditions, the new criterion is shown to be strongly consistent in the sense that with probability one, for all large n, the criterion chooses the smallest true model. The penalty function denoted by C_n depends on the sample size n and is chosen to ensure the consistency in the selection of true model. There are various choices of C_n suggested in the literature on model selection. In this paper we show that a particular choice of C_n based on observed data, which makes it random, preserves the consistency property and provides improved performance over a fixed choice of C_n.     

Computationally efficient classes of higher-order kernel functions

Classes of higher-order kernels for estimation of a probability density are constructed by iterating the twicing procedure. Given a kernel K of order l, we build a family of kernels K_m of orders l(m + 1) with the attractive property that their Fourier transforms are simply 1 — {1 —$(.)}^m+1, where ? is the Fourier transform of K. These families of higher-order kernels are well suited when the fast Fourier transform is used to speed up the calculation of the kernel estimate or the least-squares cross-validation procedure for selection of the window width. We also compare the theoretical performance of the optimal polynomial-based kernels with that of the iterative twicing kernels constructed from some popular second-order kernels.     

Variable selection in partial linear regression with functional covariate

The problem of variable selection is considered in high-dimensional partial linear regression under some model allowing for possibly functional variable. The procedure studied is that of nonconcave-penalized least squares. It is shown the existence of a √n/s_n-consistent estimator for the vector of p_n linear parameters in the model, even when p_n tends to ∞ as the sample size n increases (s_n denotes the number of influential variables). An oracle property is also obtained for the variable selection method, and the nonparametric rate of convergence is stated for the estimator of the nonlinear functional component of the model. Finally, a simulation study illustrates the finite sample size performance of our procedure.     

A novel Hybrid RBF Neural Networks model as a forecaster

We introduce a novel predictive statistical modeling technique called Hybrid Radial Basis Function Neural Networks (HRBF-NN) as a forecaster. HRBF-NN is a flexible forecasting technique that integrates regression trees, ridge regression, with radial basis function (RBF) neural networks (NN). We develop a new computational procedure using model selection based on information-theoretic principles as the fitness function using the genetic algorithm (GA) to carry out subset selection of best predictors. Due to the dynamic and chaotic nature of the underlying stock market process, as is well known, the task of generating economically useful stock market forecasts is difficult, if not impossible. HRBF-NN is well suited for modeling complex non-linear relationships and dependencies between the stock indices. We propose HRBF-NN as our forecaster and a predictive modeling tool to study the daily movements of stock indices. We show numerical examples to determine a predictive relationship between the Istanbul Stock Exchange National 100 Index (ISE100) and seven other international stock market indices. We select the best subset of predictors by minimizing the information complexity (ICOMP) criterion as the fitness function within the GA. Using the best subset of variables we construct out-of-sample forecasts for the ISE100 index to determine the daily directional movements. Our results obtained demonstrate the utility and the flexibility of HRBF-NN as a clever predictive modeling tool for highly dependent and nonlinear data.     

Bayesian Lasso-mixed quantile regression

In this paper, we discuss the regularization in linear-mixed quantile regression. A hierarchical Bayesian model is used to shrink the fixed and random effects towards the common population values by introducing an l₁ penalty in the mixed quantile regression check function. A Gibbs sampler is developed to simulate the parameters from the posterior distributions. Through simulation studies and analysis of an age-related macular degeneration (ARMD) data, we assess the performance of the proposed method. The simulation studies and the ARMD data analysis indicate that the proposed method performs well in comparison with the other approaches.     

A stochastic representation for the lp-norm symmetric distribution and its applications

The family of l_p-norm symmetric distributions was proposed by Yue and Ma and is a natural generalization to the family of l₁-norm symmetric distributions studied by Fang et al. In this article, we propose a stochastic representation for the l_p-norm symmetric distribution for any constant p > 0. The stochastic representation is expressed through independent and identically distributed uniform U(0, 1) random variables. It is illustrated that the stochastic representation can be applied to statistical simulation and uniform experimental design.