# Identifiability and estimation of recursive max-linear models

## TL;DR

This paper investigates the identifiability and estimation methods for recursive max-linear models on DAGs, providing new techniques for structure learning and parameter estimation despite non-identifiability issues.

## Contribution

It characterizes the class of DAGs and edge weights compatible with observed distributions and proposes a generalized MLE and a structure identification method.

## Key findings

- The class of DAGs and edge weights can be fully identified from observational data.
- A generalized maximum likelihood estimator for edge weights is developed.
- A simple method reliably identifies DAG structures with high probability.

## 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.