Objective This study aims to explore the impact of different prior distributions on the estimation of key parameters in Bayesian hierarchical models and to analyze how changes in the number of group levels affect estimation accuracy.Method Through case analysis and simulation studies,we assigned inverse gamma,half-Cauchy,and exponential distributions to the group variance parameters in the model,comparing their performance and the influence of the number of group levels.Results The exponential distribution,due to its shorter tail,may underestimate variance,while the inverse gamma and half-Cauchy distributions,with their thicker tails,provide more robust estimates.In scenarios with a higher number of group levels,the model estimates variance more accurately,but as the number of groups decreases,estimation errors increase.Conclusion In small sample conditions,the half-Cauchy and exponential distributions are more suitable choices because they easily incorporate external information and reasonably reflect the variation characteristics of parameters.These findings provide important guidance for selecting priors for group variance parameters in practical applications of Bayesian hierarchical models.