Comparison of Prior Setting Methods for Multilevel Model Effect Estimation Based on Small Sample Imbalanced Nested Data in Bayesian Framework

Table 4

Summary of estimation accuracy of treatment effects and nesting effects for the SSIND under all operating conditions.

	Prior setting	Convergence rate			Cover			Bias			RMSE
		Treatment effect	Nest effect		Treatment effect	Nest effect		Treatment effect	Nest effect		Treatment effect	Nest effect
Mean value	U	97.0		99.0	96.81		93.22	−101	485		79	65
	G	97.4		97.8	93.04		89.49	−69	−143		79	30
Median value	U	96.9		99.0	96.51		94.34	−76	357		77	47
	G	97.5		97.8	92.90		93.14	−56	−147		78	29
Minimum value	U	96.0		98.1	94.45		80.16	−301	−291		45	26
	G	96.3		96.7	91.27		57.58	−171	−398		46	25
Maximum value	U	98.2		100.0	99.17		97.46	−23	1979		123	198
	G	98.2		99.1	95.17		97.96	−6	59		125	40

Note. “G” means gamma prior, and “U” means uniform prior.