Validation of causal inference data using DirectLiNGAM in an environmental small-scale model and calculation settings

Area of Science:

• Environmental Science
• Data Science
• Causal Inference

Background:

• Environmental fields like marine, weather, and soil data often involve small datasets with limited analytical scope.
• Existing statistical methods for factor association in small datasets lack consistent approaches for group assessment.

Purpose of the Study:

• To present essential checkpoints and settings for calculating linear non-Gaussian structural equation models (DirectLiNGAM).
• To describe validation methods for DirectLiNGAM results using small-scale model data.
• To provide statistical validation for association networks, treatments, and interventions in structural inference.

Main Methods:

• Utilizing causal inference with the DirectLiNGAM method.
• Validating results through correlation coefficient and feature importance analysis.
• Employing causal effect objects and propensity scores for validation.

Main Results:

• DirectLiNGAM offers effective results for small data, identifying potential associations between factors.
• Statistical validation methods are provided for assessing group associations.
• The study discusses checkpoints and settings for DirectLiNGAM calculations.

Conclusions:

• Causal inference with DirectLiNGAM is a promising approach for small environmental datasets.
• The proposed validation methods enhance the reliability of factor association analysis.
• This work contributes to understanding potential associations and structural inference in environmental data analysis.