Objective A scientific and effective method for analyzing haze influencing factors is established to solve the regional moderate and mild air pollution in the Chengdu-Chongqing area. Methods The hourly monitoring data of air quality from 2014 to 2022 in Chengdu-Chongqing Economic Circle were obtained and the nested structure characteristics of the monitoring data were used to construct a three-level Bayesian development model at the annual time-site-region level and the comparative analysis of classical OLS IGLS estimation empirical Bayesian and complete Bayesian was carried out which demonstrated the advantages of the complete Bayesian method. Results The PM2. 5 concentration is affected by comprehensive factors including air pollutants meteorology population economy industrial structure and agricultural production and the change rate of PM2. 5 concentration is affected by CO SO2 and urbanization rate. Conclusion The use of hierarchical modeling to analyze haze monitoring data with nested multi-level structures is more scientific and reasonable. By incorporating Bayesian statistics which allows for the utilization of prior information and learning mechanisms it can further enhance the accuracy of model parameter estimation and prediction.