Skip to main content
. Author manuscript; available in PMC: 2014 Jul 1.
Published in final edited form as: Multivariate Behav Res. 2013 Jul 26;48(4):563–591. doi: 10.1080/00273171.2013.802647

Table 4.

ML and Bayesian Estimation Performances for Multilevel MIRT Models

Parameters vs. ML estimates Parameters vs. Bayesian estimates

Factors Items Covariates Sample size Parameters AAB RMSE AAB RMSE
2 3 0 500 16 0.062 0.086 0.102 0.137
1,000 16 0.030 0.048 0.052 0.069
3,000 16 0.026 0.042 0.023 0.030

1 500 18 0.039 0.055 0.078 0.106
1,000 18 0.025 0.041 0.043 0.056
3,000 18 0.022 0.040 0.013 0.017

6 0 500 28 0.045 0.059 0.107 0.141
1,000 28 0.031 0.043 0.052 0.068
3,000 28 0.025 0.039 0.020 0.026

1 500 30 0.033 0.046 0.069 0.094
1,000 30 0.024 0.038 0.034 0.046
3,000 30 0.022 0.037 0.012 0.017

3 3 0 500 25 0.053 0.076 0.141 0.187
1,000 25 0.029 0.039 0.078 0.102
3,000 25 0.025 0.033 0.027 0.034

1 500 28 0.059 0.112 0.089 0.133
1,000 28 0.030 0.046 0.048 0.071
3,000 28 0.023 0.032 0.017 0.026

6 0 500 43 0.047 0.071 0.116 0.151
1,000 43 0.039 0.061 0.057 0.076
3,000 43 0.031 0.048 0.015 0.021

1 500 46 0.034 0.051 0.094 0.128
1,000 46 0.029 0.044 0.049 0.066
3,000 46 0.023 0.033 0.017 0.024

Note. AAB represents averaged absolute bias; RMSE represents root mean square error.