Skip to main content
. Author manuscript; available in PMC: 2009 Oct 1.
Published in final edited form as: J Am Stat Assoc. 2008 Mar 1;103(481):61–73. doi: 10.1198/016214507000000329

Table 8.

Simulations with a common sensitivity and specificity

r FM model average estimate GRE model average estimate % (log LGRE ≫ log LFM)*
Pd SENS SPEC Pd SENS SPEC
0 .24 (.03) .84 (.04) .95 (.01) .20 (.06) .73 (.18) .88 (.06) 0
.2 .22 (.01) .81 (.04) .92 (.01) .20 (.02) .75 (.04) .90 (.01) 100
.5 .21 (.01) .78 (.03) .91 (.01) .20 (.02) .75 (.03) .90 (.01) 100
.8 .21 (.01) .76 (.03) .90 (.01) .20 (.01) .75 (.03) .90 (.01) 100
1 .20 (.01) .75 (.02) .90 (.01) .20 (.01) .75 (.02) .90 (.01) 100

NOTE: Data were simulated under the Gaussian random effects (GRE) model with Pd = .2, σ0 = σ1 = 1.5, SENS = .75, SPEC = .90, I = 1,000, and J = 5. Results are based on 1,000 simulations.

*

Proportion of realizations where the log-likelihood under the GRE model is more than 1 larger than the log-likelihood under the misspecified FM model.