Causal Analysis of Stunting Determinants Using the Peter-Clark and Greedy Equivalence Search Algorithms

Abstract

Child stunting remains a major public health challenge, reflecting the long-term effects of inadequate nutrition, limited maternal education, and poor access to health services. Understanding the causal structure underlying these factors is essential to design effective interventions. This study employs two causal discovery algorithms Greedy Equivalence Search (GES) and Peter Clark (PC) to analyze the causal relationships among key determinants of stunting using secondary data from the West Bangka District Health Office (2024). The dataset includes eight variables related to anthropometric measurements, maternal characteristics, and environmental conditions. Model performance was evaluated using Directed Density (DD) and Causal Density (CD) metrics to measure the strength and sparsity of the causal networks. The GES algorithm produced a well-structured causal model showing that maternal education, posyandu (community health post) visits, and exclusive breastfeeding were primary causal drivers influencing height-for-age (TB/U) and weight-for-age (BB/U). The PC algorithm, based on conditional independence testing, revealed a similar but less dense causal network, identifying only direct and statistically robust associations. Evaluation results indicated that the PC model achieved a Directed Density (DD) of 0.66 and a Causal Density (CD) of 0.07, while the GES model showed higher directedness and complexity in its causal mapping. Both algorithms successfully identified meaningful causal structures among stunting-related variables, with GES capturing broader causal pathways and PC emphasizing stronger direct causal links. These findings demonstrate that integrating score-based and constraint-based causal models provides complementary insights into the mechanisms driving stunting.

GS Cited Analysis