Structured Additive Distributional Regression

结构化加性分布回归

• 批准号: 166547046
• 负责人: Professor Dr. Thomas Kneib
• 金额: --
• 依托单位: Lehrstuhl für Statistik
• 依托单位国家: 德国
• 项目类别: Research Grants
• 财政年份: 2010
• 资助国家: 德国
• 起止时间: 2009-12-31 至 2015-12-31
• 项目状态: 已结题
• 来源: https://gepris.dfg.de/gepris/projekt/166547046?language=en
• 关键词: Structured; Additive; Distributional; Regression

Regression models form one of the standard tools for empirical analyses in all scientific disciplines and in particular in economics. While usual regression specifications such as the linear model or generalized linear models aim at describing the expectation of a response conditional on covariates, recent interest has shifted towards models that allow to analyse more general properties of the distribution of a response (we will refer to this as distributional regression in the following). This comprises completely distribution free approaches such as quantile and expectile regression as well as flexible parametric approaches for location, scale and shape. The increasing availability of complex covariate information also induces a requirement for similarly complex predictor specifications such as nonlinear and spatial effects in the context of distributional regression.This project extends different classes of distributional regression and develops corresponding inferential techniques. The model classes considered comprise different versions of quantile and expectile regression, modal regression and regression models for location, scale and shape. The developed methods will furthermore be employed in the different areas of application for empirical analyses.

回归模型是所有科学领域，特别是经济学领域进行实证分析的标准工具之一。虽然通常的回归规范，如线性模型或广义线性模型，旨在描述以协变量为条件的响应的预期，但最近的兴趣已转向允许分析响应分布的更一般性质的模型(我们将在下文中将其称为分布回归)。这包括完全无分布的方法，如分位数和预期回归，以及位置、规模和形状的灵活参数方法。复杂协变量信息的增加也导致了对类似复杂预测指标的需求，例如分布回归中的非线性和空间效应。本项目扩展了不同类型的分布回归，并开发了相应的推理技术。所考虑的模型类别包括不同版本的分位数和预期回归、模式回归和位置、规模和形状的回归模型。所开发的方法还将用于不同的应用领域进行实证分析。

项目成果

Modelling Hospital Admission and Length of Stay by Means of Generalised Count Data Models

通过广义计数数据模型对入院和住院时间进行建模

• DOI: 10.1002/jae.2454
• 发表时间: 2016
• 期刊: Journal of Applied Econometrics
• 影响因子: 2.1
• 作者: Herwartz;Strumann
• 通讯作者: Strumann

Simultaneous inference in structured additive conditional copula regression models: a unifying Bayesian approach

• DOI: 10.1007/s11222-015-9573-6
• 发表时间: 2016-07
• 期刊: Statistics and Computing
• 影响因子: 2.2
• 作者: N. Klein;T. Kneib

Scale-Dependent Priors for Variance Parameters in Structured Additive Distributional Regression

结构化加性分布回归中方差参数的尺度相关先验

• DOI: 10.1214/15-ba983
• 发表时间: 2016
• 期刊: Bayesian Analysis
• 影响因子: 4.4

Bayesian Generalized Additive Models for Location, Scale, and Shape for Zero-Inflated and Overdispersed Count Data

• DOI: 10.1080/01621459.2014.912955
• 发表时间: 2015-03-01
• 期刊: JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION
• 影响因子: 3.7
• 作者: Klein, Nadja;Kneib, Thomas;Lang, Stefan
• 通讯作者: Lang, Stefan

BAYESIAN STRUCTURED ADDITIVE DISTRIBUTIONAL REGRESSION WITH AN APPLICATION TO REGIONAL INCOME INEQUALITY IN GERMANY

• DOI: 10.1214/15-aoas823
• 发表时间: 2015-06-01
• 期刊: ANNALS OF APPLIED STATISTICS
• 影响因子: 1.8
• 作者: Klein, Nadja;Kneib, Thomas;Sohn, Alexander
• 通讯作者: Sohn, Alexander