Skip to main content
. Author manuscript; available in PMC: 2016 Sep 7.
Published in final edited form as: Biometrics. 2016 Feb 17;72(3):986–994. doi: 10.1111/biom.12492

Table 2. Simulation results for the mean ratio (one unit increase in continuous exposure), for the derived mean estimation from a latent (conventional) log-logit model, and the direct mean estimation from a marginal log-logit working model, with data generated from ZINB models.

Latent log-logit working model
Derived marginal mean estimation
Marginal log-logit working model
Direct marginal mean estimation


True MR n EMR RB(%) MSE CP(%) AIC EMR RB(%) MSE CP(%) AIC
Data generated from a ZINB model with a marginally specified mean
0.607 50 0.592 -2.38 0.014 97.4 162.16 0.620 2.22 0.012 94.3 162.29
100 0.592 -2.41 0.007 98.5 317.52 0.612 0.94 0.005 93.4 317.34
200 0.592 -2.32 0.003 99.3 625.57 0.606 -0.05 0.003 94.6 625.31
500 0.594 -2.10 0.001 98.8 1559.95 0.605 -0.30 0.001 94.6 1559.14
1000 0.595 -1.88 0.0007 98.7 3110.13 0.606 -0.11 0.0005 93.6 3108.68