We consider the generic regularized optimization problem ˆ β(λ) = arg minβ L(y, Xβ) + λJ(β). Recently, Efron et al. (2004) have shown that for the Lasso – that is, if L is squared error loss and J(β) = ‖β‖1 is the l1 norm of β – the optimal coefficient path is piecewise linear, i.e., ∂ ˆ β(λ)/∂λ is piecewise constant. We derive a general characterization of the properties of (loss L, penalty J) pairs which give piecewise linear coefficient paths. Such pairs allow for efficient generation of the full regularized coefficient paths. We investigate the nature of efficient path following algorithms which arise. We use our results to suggest robust versions of the Lasso for regression and classification, and to develop new, efficient algorithms for existing problems in the literature, including Mammen & van de Geer’s Locally Adaptive Regression Splines. 1

Variable selection plays an important role in high dimensional statistical modeling which nowa-days appears in many areas and is key to various scientific discoveries. For problems of large scale or dimensionality p, estimation accuracy and computational cost are two top concerns. In a recent paper, Candes and Tao (2007) propose the Dantzig selector using L1 regularization and show that it achieves the ideal risk up to a logarithmic factor log p. Their innovative procedure and remarkable result are challenged when the dimensionality is ultra high as the factor log p can be large and their uniform uncertainty principle can fail. Motivated by these concerns, we introduce the concept of sure screening and propose a sure screening method based on a correlation learning, called the Sure Independence Screening (SIS), to reduce dimensionality from high to a moderate scale that is below sample size. In a fairly general asymptotic framework, the SIS is shown to have the sure screening property for even exponentially growing dimensionality. As a methodological extension, an iterative SIS (ISIS) is also proposed to enhance its finite sample performance. With dimension reduced accurately from high to below sample size, variable selection can be improved on both speed and accuracy, and can then be ac-

Abstract. Technological innovations have revolutionized the process of scientific research and knowledge discovery. The availability of massive data and challenges from frontiers of research and development have reshaped statistical thinking, data analysis and theoretical studies. The challenges of high-dimensionality arise in diverse fields of sciences and the humanities, ranging from computational biology and health studies to financial engineering and risk management. In all of these fields, variable selection and feature extraction are crucial for knowledge discovery. We first give a comprehensive overview of statistical challenges with high dimensionality in these diverse disciplines. We then approach the problem of variable selection and feature extraction using a unified framework: penalized likelihood methods. Issues relevant to the choice of penalty functions are addressed. We demonstrate that for a host of statistical problems, as long as the dimensionality is not excessively large, we can estimate the model parameters as well as if the best model is known in advance. The persistence property in risk minimization is also addressed. The applicability of such a theory and method to diverse statistical problems is demonstrated. Other related problems with high-dimensionality are also discussed.

Variable selection is fundamental to high-dimensional statistical modeling. Many variable selection techniques may be implemented by maximum penalized likelihood using various penalty functions. Optimizing the penalized likelihood function is often challenging because it may be nondifferentiable and/or nonconcave. This article proposes a new class of algorithms for finding a maximizer of the penalized likelihood for a broad class of penalty functions. These algorithms operate by perturbing the penalty function slightly to render it differentiable, then optimizing this differentiable function using a minorize–maximize (MM) algorithm. MM algorithms are useful extensions of the well-known class of EM algorithms, a fact that allows us to analyze the local and global convergence of the proposed algorithm using some of the techniques employed for EM algorithms. In particular, we prove that when our MM algorithms converge, they must converge to a desirable point; we also discuss conditions under which this convergence may be guaranteed. We exploit the Newton–Raphson-like aspect of these algorithms

Summary. We study the asymptotic properties of adaptive LASSO estimators in sparse, high-dimensional, linear regression models when the number of covariates may increase with the sample size. We consider variable selection using the adaptive LASSO, where the L1 norms in the penalty are re-weighted by data-dependent weights. We show that, if a reasonable initial estimator is available, then under appropriate conditions, adaptive LASSO correctly select covariates with nonzero coefficients with probability converging to one and that the estimators of nonzero coefficients have the same asymptotic dis-tribution that they would have if the zero coefficients were known in advance. Thus, the adaptive LASSO has an oracle property in the sense of Fan and Li (2001) and Fan and Peng (2004). In addition, under a partial orthogonality condition in which the covariates with zero coefficients are weakly correlated with the covariates with nonzero coefficients, univariate regression can be used to obtain the initial estimator. With this initial estimator, adaptive LASSO has the oracle property even when the number of covariates is greater than the sample size. Key Words and phrases. Penalized regression, high-dimensional data, variable selection, asymptotic normality, oracle property, zero-consistency. Short title. Sparse high-dimensional regression

Summary. We study the asymptotic properties of bridge estimators in sparse, high-dimensional, linear regression models when the number of covariates may increase to infinity with the sample size. We are particularly interested in the use of bridge estima-tors to distinguish between covariates whose coefficients are zero and covariates whose coefficients are nonzero. We show that under appropriate conditions, bridge estimators correctly select covariates with nonzero coefficients with probability converging to one and that the estimators of nonzero coefficients have the same asymptotic distribution that they would have if the zero coefficients were known in advance. Thus, bridge es-timators have an oracle property in the sense of Fan and Li (2001) and Fan and Peng (2004). In general, the oracle property holds only if the number of covariates is smaller than the sample size. However, under a partial orthogonality condition in which the co-variates of the zero coefficients are uncorrelated or weakly correlated with the covariates of nonzero coefficients, we show that marginal bridge estimators can correctly distin-guish between covariates with nonzero and zero coefficients with probability converging to one even when the number of covariates is greater than the sample size. Key Words and phrases. Penalized regression, high-dimensional data, variable selection, asymptotic normality, oracle property. Short title. Sparse high-dimensional regression

We derive conditions guaranteeing estimation and model selection consistency, oracle properties and persistence for the group-lasso estimator and model selector proposed by Yuan and Lin (2006) for least squares problems when the covariates have a natural grouping structure. We study both the case of a fixed-dimensional parameter space with increasing sample size and the case when the model complexity changes with the sample size. 1

The Lasso is an attractive regularisation method for high dimensional regression. It combines variable selection with an efficient computational procedure. However, the rate of convergence of the Lasso is slow for some sparse high dimensional data, where the number of predictor variables is growing fast with the number of observations. Moreover, many noise variables are selected if the estimator is chosen by cross-validation. It is shown that the contradicting demands of an efficient computational procedure and fast convergence rates of the ℓ2-loss can be overcome by a two-stage procedure, termed the relaxed Lasso. For orthogonal designs, the relaxed Lasso provides a continuum of solutions that include both soft- and hard-thresholding of estimators. The relaxed Lasso solutions include all regular Lasso solutions and computation of all relaxed Lasso solutions is often identically expensive as computing all regular Lasso solutions. Theoretical and numerical results demonstrate that the relaxed Lasso produces sparser models with equal or lower prediction loss than the regular Lasso estimator for highdimensional data. 1

The use of principal component methods to analyze functional data is appropriate in a wide range of different settings. In studies of “functional data analysis, ” it has often been assumed that a sample of random functions is observed precisely, in the continuum and without noise. While this has been the traditional setting for functional data analysis, in the context of longitudinal data analysis a random function typically represents a patient, or subject, who is observed at only a small number of randomly distributed points, with nonnegligible measurement error. Nevertheless, essentially the same methods can be used in both these cases, as well as in the vast number of settings that lie between them. How is performance affected by the sampling plan? In this paper we answer that question. We show that if there is a sample of n functions, or subjects, then estimation of eigenvalues is a semiparametric problem, with root-n consistent estimators, even if only a few observations are made of each function,

Recent results in homotopy and solution paths demonstrate that certain well-designed greedy algorithms, with a range of values of the algorithmic parameter, can provide solution paths to a sequence of convex optimization problems. On the other hand, in regression many existing criteria in subset selection (including Cp, AIC, BIC, MDL, RIC, etc.) involve optimizing an objective function that contains a counting measure. The two optimization problems are formulated as (P1) and(P0) in the present paper. The latter is generally combinatoric and has been proven to be NP-hard. We study the conditions under which the two optimization problems have common solutions. Hence, in these situations a stepwise algorithm can be used to solve the seemingly unsolvable problem. Our main result is motivated by recent work in sparse representation, while two others emerge from different angles: a direct analysis of sufficiency and necessity and a condition on the mostly correlated covariates. An extreme example connected with least angle regression is of independent interest. 1. Introduction. We