Development of uncertainty-based work injury model using Bayesian structural equation modelling.

Area of Science:

• Occupational Health and Safety
• Statistical Modeling
• Mining Engineering

Background:

• Miners' work injury poses significant risks in underground coal mines.
• Accurate modeling of work injury is crucial for implementing effective safety measures.
• Traditional SEM methods may not fully capture the complexities of injury causation.

Purpose of the Study:

• To propose a Bayesian structural equation model (SEM) for analyzing miners' work injury in an Indian underground coal mine.
• To compare the performance of expert-opinion-based priors versus fixed priors in Bayesian SEM for work injury.
• To identify key environmental and behavioral variables influencing work injury.

Main Methods:

• Developed a Bayesian structural equation model (SEM) incorporating environmental and behavioral variables.
• Employed two approaches for prior distributions: fixed distributions and expert opinions.
• Utilized Bayesian inference and Markov Chain Monte Carlo (MCMC) sampling (Gibbs sampling) to obtain posterior distributions.
• Evaluated model fit using coefficient of determination and mean squared error.

Main Results:

• Bayesian SEM with expert-opinion-based priors showed all structural and measurement model coefficients were statistically significant.
• Fixed prior distributions resulted in two non-significant coefficients.
• The Bayesian structural model demonstrated a good fit for work injury data (R-squared = 0.91) with lower mean squared error than traditional SEM.

Conclusions:

• Expert-opinion-based priors enhance the statistical significance and reliability of Bayesian SEM for analyzing miners' work injury.
• The proposed Bayesian SEM offers a more accurate and robust approach to understanding and mitigating work injury risks in mining environments.
• This methodology provides valuable insights for improving safety protocols in underground coal mines.