Skip to main content

Table 2 Posterior mean of the deviance ( D ̄ ), deviance of the mean ( D ( θ ̄ ) ), effective number of parameters ( p D ) as measure of model complexity, DIC value, and mean of logarithmic scores ( LS ¯ )

From: Bayesian model selection techniques as decision support for shaping a statistical analysis plan of a clinical trial: An example from a vertigo phase III study with longitudinal count data as primary endpoint

Model

D Ì„

D ( θ ̄ )

p D

DIC

LS ¯

p-value†

Poisson, I

2034

1931

103

2137

2.046

<0.0001

Poisson, IS

1633

1463

170

1803

1.641

ref.

ZIP, I

2020

1917

103

2123

2.018

<0.0001

ZIP, IS

1637

1467

170

1807

1.644

0.5797

NB, I

2006

1903

103

2109

1.932

<0.0001

NB, IS

1659

1489

170

1829

1.665

0.6841

arcsinh, I

994∗

886∗

108∗

1102∗

1.967

n.a.

arcsinh, IS

834∗

636∗

198∗

1032∗

1.960

n.a.

  1. ∗ Comparison of DIC for NMMs is not applicable because of different arcsinh-transformed outcomes.
  2. †Monte-Carlo permutation test for paired individual logarithmic scores.