Identifiability and estimation of recursive max-linear models

DOI10.1111/SJOS.12446zbMATH Open1467.62105arXiv1901.03556OpenAlexW3023983480MaRDI QIDQ4994811FDOQ4994811

Authors: Nadine Gissibl, Claudia Klüppelberg, Steffen Lauritzen 

Publication date: 22 June 2021

Published in: Scandinavian Journal of Statistics (Search for Journal in Brave)

Abstract: We address the identifiablity and estimation of recursive max-linear structural equation models represented by an edge weighted directed acyclic graph (DAG). Such models are generally unidentifiable and we identify the whole class of DAGs and edge weights corresponding to a given observational distribution. For estimation, standard likelihood theory cannot be applied because the corresponding families of distributions are not dominated. Given the underlying DAG, we present an estimator for the class of edge weights and show that it can be considered a generalized maximum likelihood estimator. In addition, we develop a simple method for identifying the structures of the DAGs. With probability tending to one at an exponential rate with the number of observations, this method correctly identifies the class of DAGs and, similarly, exactly identifies the possible edge weights.

Full work available at URL: https://arxiv.org/abs/1901.03556

Recommendations

• Max-linear models on directed acyclic graphs
• Recursive max-linear models with propagating noise
• Estimating an extreme Bayesian network via scalings
• Computing maximum likelihood estimates in recursive linear models with correlated errors
• Discrete max-linear Bayesian networks

zbMATH Keywords

causal inferencestructural equation modelBayesian networkextreme value theorygraphical modeldirected acyclic graph (DAG)generalized maximum likelihood estimation

Mathematics Subject Classification ID

Applications of graph theory (05C90) Statistics of extreme values; tail inference (62G32) Probabilistic graphical models (62H22)

Cited In (11)

• Causal discovery in heavy-tailed models
• Estimating an extreme Bayesian network via scalings
• Max-linear models on directed acyclic graphs
• Recursive max-linear models with propagating noise
• Max-linear models in random environment
• Computing maximum likelihood estimates in recursive linear models with correlated errors
• Max-linear regression models with regularization
• Recursive maximum likelihood identification of a non-linear output-affine model
• Discrete max-linear Bayesian networks
• Conditional independence in max-linear Bayesian networks
• Max-linear graphical models with heavy-tailed factors on trees of transitive tournaments

This page was built for publication: Identifiability and estimation of recursive max-linear models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4994811)