NMA Model
|
Based on |
Generalized Linear Models [6], [17]
|
Multivariate methods [18]
|
Graph theory [32]
|
Model options regarding the Homogeneity assumption
|
Fixed-effect model |
✓ |
✓ |
✓ |
|
Random-effect model with a common heterogeneity parameter |
✓ |
✓ |
✓ |
|
Random-effect model with different heterogeneity parameters |
✗ |
✓ |
✗ |
Model options regarding the Consistency assumption
|
Consistency model |
✓ |
✓ |
✓ |
|
Inconsistency model |
✓ |
✗ |
✓ |
Inclusion of covariates
|
Meta-regression |
✗ |
✗ |
✗ |
Estimation framework
|
Frequentist |
✗ |
✗ |
✓ |
|
Bayesian |
✓ |
✓ |
✗ |
Bayesian Modeling
|
NA |
Prior distributions for baseline and relative effect parameters
|
Default distribution and parameter values |
✓ Normal distribution, heuristic initial values |
✓ Normal distribution, heuristic initial values |
|
|
Option for user-specified distribution and parameter values |
✓ Restricted to specifying variance |
✓ Restricted to Normal distribution |
|
Prior distribution for heterogeneity parameters
|
Default distribution and parameter values |
✓ Uniform distribution, heuristic initial values |
✓ Inverse-Gamma distribution, specific values |
|
|
Option for user-specified distribution and parameter values |
✓ Uniform or Gamma distribution, specify values |
✓ Inverse-Gamma or Wishart distribution, specify values |
|
Markov Chain Monte Carlo (MCMC) Sampler
|
WinBUGS |
✓ |
✗ |
|
|
OpenBUGS |
✓ |
✗ |
|
|
JAGS |
✓ |
✓ |
|
Control over posterior samples
|
Total iterations |
✓ |
✓ |
|
|
Adaptation phase |
✓ |
✓ |
|
|
Burn-in phase |
✓ |
✓ |
|
|
Thinning |
✓ |
✓ |
|
Model convergence diagnostics
|
Option for multiple chains |
✓ |
✓ |
|
|
Time-series plot |
✓ |
✓ |
|
|
Trace plot |
✓ |
✗ |
|
|
Brooks-Gelman-Rubin (BGR) diagnostic test |
✓ |
✗ |
|