論文

We consider the problem of estimating sparse vector autoregression (VAR) via penalized precision matrices. This matrix is the output of the underlying directed acyclic graph of the VAR process, whose zero components correspond to the zero coefficients of the graphical representation of the VAR. The sparsity-based precision matrix estimator is deduced from the D-trace loss with convex and nonconvex penalty functions. We establish the consistency of the penalized estimator and provide the conditions for which all true zero entries of the precision matrix are actually estimated as zero with probability tending to one. The relevance of the method is supported by simulated experiments and a real data application.

This article analyses three feature screening procedures: Kendall's Tau and Spearman Rho (TR), Hilbert-Schmidt Independence Criterion (HSIC) and conditional Maximum Mean Discrepancy (cMMD), where the latter is a modified version of the standard MMD for categorical classification. These association measures are not based on any specific underlying model, such as the linear regression. We provide the conditions for which the sure independence screening (SIS) property is satisfied under a lower bound assumption on the minimum signal strength of the association measure. The SIS property for the HSIC and cMMD is established for given bounded and symmetric kernels. Within the high-dimensional setting, we propose a two-step approach to control the false discovery rate (FDR) using the knockoff filtering. The performances of the association measures are assessed through simulated and real data experiments and compared with existing competing screening methods.

We introduce a general methodology to consistently estimate multidimensional ARCH models equation-by-equation, possibly with a very large number of parameters through penalization (Sparse Group Lasso). Some families of multidimensional ARCH models are proposed to tackle homogeneous or heterogeneous portfolios of assets. The corresponding conditions of stationarity and of positive definiteness are studied. We evaluate the relevance of such a strategy by simulation. The relative forecasting performances of our models are compared through the management of financial portfolios.

Although multivariate stochastic volatility models usually produce more accurate forecasts compared with the MGARCH models, their estimation techniques such as Bayesian MCMC typically suffer from the curse of dimensionality. We propose a fast and efficient estimation approach for MSV based on a penalized OLS framework. Specifying the MSV model as a multivariate state-space model, we carry out a two-step penalized procedure. We provide the asymptotic properties of the two-step estimator and the oracle property of the first-step estimator when the number of parameters diverges. The performances of our method are illustrated through simulations and financial data. Supplementary Material presenting technical proofs is available online.

This article is written in memory of Harry Markowitz, the founder of modern portfolio theory. We report a few human perspectives of his character, we review a large number of his contributions, published both in operations research and finance oriented journals, and we focus on one of the most critical, and still open, portfolio theory issues, the forecast of covariance matrices. Our contribution in this paper is placed exactly towards this direction. More specifically, we compare the performances of several approaches to predict the variance-covariance matrices of vectors of asset returns, through simulated and real data experiments: some dynamic models such as Dynamic Conditional Correlation (DCC) and C-vine GARCH on one side, and several agnostic methods (Average Oracle, usual “Sample” matrix) on the other side. The most robust methods seem to be DCC and the Average Oracle approaches.

Detecting influential features in non-linear and/or high-dimensional data is a challenging and increasingly important task in machine learning. Variable selection methods have thus been gaining much attention as well as post-selection inference. Indeed, the selected features can be significantly flawed when the selection procedure is not accounted for. We propose a selective inference procedure using the so-called model-free”HSIC-Lasso” based on the framework of truncated Gaussians combined with the polyhedral lemma. We then develop an algorithm, which allows for low computational costs and provides a selection of the regularisation parameter. The performance of our method is illustrated by both artificial and real-world data based experiments, which emphasise a tight control of the type-I error, even for small sample sizes.

We study the large-sample properties of sparse M-estimators in the presence of pseudo-observations. Our frame work covers a broad class of semi-parametric copula models, for which the marginal distributions are unknown and replaced by their empirical counterparts. It is well known that the latter modification significantly alters the limiting laws compared to usual M-estimation. We establish the consistency and the asymptotic normality of our sparsepenalized M-estimator and we prove the asymptotic oracle property with pseudo-observations, possibly in the case when the number of parameters is diverging. Our framework allows to manage copula-based loss functions that are potentially unbounded. Additionally, we state the weak limit of multivariate rank statistics for an arbitrary dimension and the weak convergence of empirical copula processes indexed by maps. We apply our inference method to Canonical Maximum Likelihood losses with Gaussian copulas, mixtures of copulas or conditional copulas. The theoretical results are illustrated by two numerical experiments.

We provide finite sample properties of general regularized statistical criteria in the presence of pseudo-observations. Under the restricted strong convexity assumption of the unpenalized loss function and regularity conditions on the penalty, we derive non-asymptotic error bounds on the regularized M-estimator. This penalized framework with pseudo-observations is then applied to the M-estimation of some usual copula-based models. These theoretical results are supported by an empirical study.