---
title: "A Study of Causal Discovery With Weak Links and Small Samples"
authors:
  - "Honghua Dai"
  - "Kevin B. Korb"
  - "Chris S. Wallace"
  - "Xindong Wu"
doi: ""
url: "https://ijcai.org/proceedings/1997/188"
pdf: "https://ijcai.org/proceedings/1997/0188.pdf"
venue: "Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence"
year: 1997
pages: "1304-1308"
track: "Probabilistic Reasoning"
keywords:
---

# A Study of Causal Discovery With Weak Links and Small Samples

Honghua Dai, Kevin B. Korb, Chris S. Wallace, Xindong Wu

*Proceedings of the Fifteenth International Joint Conference on Artificial Intelligence*, pages 1304-1308.

## Abstract

Weak causa l relationships and small sample size pose two significant difficulties to the automatic discovery of causa l models from observational data . This paper examines the influence of weak causa l links and varying sample sizes on the discovery of causa l models. The experimental results illustrate the effect of larger sample sizes for discovering causal models reliably and the relevance of the strength of causal links and the comp lexity of the original ca usal model. We present indicative evidence of th e superior robustness of MML (Minirnum M essage Length) methods to sta ndard significance t es ts in the recovery of ca usal links. The comparative results show that the MML-CI (the MML Causal Inducer) causal discovery system finds better models than TETRAD II given small samples from linear causal models. The experimental results also reveal that MML-CI finds weak links with smaller sample sizes than can TETRAD II.