A likelihood ratio framework for high-dimensional semiparametric regression

DOI10.1214/16-AOS1483MaRDI QIDQ682282zbMATH OpenFDO

Authors Yang Ning, Tianqi Zhao, Han Liu

Publication date 14 February 2018

Published in The Annals of Statistics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1412.2295, https://projecteuclid.org/euclid.aos/1513328574

zbMATH Keywords

asymptotic distributionconfidence regionsgeneralized linear modellikelihood ratio test

Mathematics Subject Classification ID

Asymptotic distribution theory in statistics (62E20) Nonparametric regression and quantile regression (62G08) Nonparametric hypothesis testing (62G10) Nonparametric tolerance and confidence regions (62G15) Asymptotic properties of nonparametric inference (62G20)

Abstract: We propose a likelihood ratio based inferential framework for high dimensional semiparametric generalized linear models. This framework addresses a variety of challenging problems in high dimensional data analysis, including incomplete data, selection bias, and heterogeneous multitask learning. Our work has three main contributions. (i) We develop a regularized statistical chromatography approach to infer the parameter of interest under the proposed semiparametric generalized linear model without the need of estimating the unknown base measure function. (ii) We propose a new framework to construct post-regularization confidence regions and tests for the low dimensional components of high dimensional parameters. Unlike existing post-regularization inferential methods, our approach is based on a novel directional likelihood. In particular, the framework naturally handles generic regularized estimators with nonconvex penalty functions and it can be used to infer least false parameters under misspecified models. (iii) We develop new concentration inequalities and normal approximation results for U-statistics with unbounded kernels, which are of independent interest. We demonstrate the consequences of the general theory by using an example of missing data problem. Extensive simulation studies and real data analysis are provided to illustrate our proposed approach.

Review text

The authors proposed a new inferential framework for high-dimensional semiparametric generalized linear models. The proposed model is closely related to the proportional likelihood ratio model (see [\textit{X. Luo} and \textit{W. Y. Tsai}, Biometrika 99, No. 1, 211--222 (2012; Zbl 1437.62545)] or [\textit{K. C. G. Chan}, Biometrika 100, No. 1, 269--276 (2013; Zbl 1452.62570)]), but this model does not require the density assumption for the nonparametric function. The proposed estimation procedure is related to the permutation based test (see [\textit{J. D. Kalbfleisch}, J. Am. Stat. Assoc. 73, 167--170 (1978; Zbl 0376.62028)]) and the second-order approximation reduces to the pairwise likelihood (see [\textit{K. C. G. Chan}, Biometrika 100, No. 1, 269--276 (2013; Zbl 1452.62570)] or [\textit{X. Luo} and \textit{W. Y. Tsai}, Biometrika 99, No. 1, 211--222 (2012; Zbl 1437.62545)]). The main results are concerning to analyze some important theoretical properties of the obtained confidence regions and establish the asymptotic distribution of the directional likelihood ratio test statistic. Some interesting numerical results are presented.

Recommendations

• A unified theory of confidence regions and testing for high-dimensional estimating equations
• High-dimensional empirical likelihood inference
• A general theory of hypothesis tests and confidence regions for sparse high dimensional models
• On High-Dimensional Constrained Maximum Likelihood Inference
• Drawing inferences for high-dimensional linear models: a selection-assisted partial regression and smoothing approach

Cited in

(8)

• A unified theory of confidence regions and testing for high-dimensional estimating equations
• A proportional likelihood ratio model
• On High-Dimensional Constrained Maximum Likelihood Inference
• Very high dimensional semiparametric models. Abstracts from the workshop held October 2--8, 2011.
• On semiparametric exponential family graphical models
• High-dimensional inference for cluster-based graphical models
• Targeted Inference Involving High-Dimensional Data Using Nuisance Penalized Regression
• Semiparametric approach to estimation of marginal mean effects and marginal quantile effects

This page was built for publication: A likelihood ratio framework for high-dimensional semiparametric regression

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q682282)