Table 4.
Slope Variance Estimate Bias by Estimation Method, Number of Clusters, and Population Slope Variance.
| MCMC |
|||||
|---|---|---|---|---|---|
| Number of clusters | Pop slope variance | REML | KR | Noninformative | Informative |
| 5 | .25 | .00129 | .00129 | .50004 | −.02523 |
| .50 | −.00611 | −.00611 | .90889 | −.02166 | |
| .75 | .00370 | .00370 | 1.30433 | −.01942 | |
| 10 | .25 | −.00011 | −.00011 | .13153 | −.01939 |
| .50 | .00109 | .00109 | .24336 | −.01598 | |
| .75 | −.00379 | −.00379 | .34483 | −.01313 | |
| 15 | .25 | .00097 | .00097 | .06809 | −.01485 |
| .50 | −.00195 | −.00195 | .11878 | −.01119 | |
| .75 | .00221 | .00221 | .17822 | −.00809 | |
| 20 | .25 | −.00031 | −.00031 | .04781 | −.01404 |
| .50 | .00047 | .00047 | .08876 | −.01222 | |
| .75 | .00106 | .00106 | .12918 | .01231 | |
Note. REML = multilevel modeling using restricted maximum likelihood estimation; KR = Kenward–Rogers correction for multilevel modeling; MCMC = Markov chain Monte Carlo with informative or noninformative priors.