Fig. 4From: Bayesian model selection for the Drosophila gap gene networkLog marginal likelihood estimates. Thermodynamic estimate of the logarithm marginal likelihood for all models. Estimates were generated for 10 independent runs for each model and show low variance. Difference between the estimates for models reveals the log Bayes factor that can be used for model comparison (see Table 2). We see that addition of a node specific-parameter for Bicoid improves the model fit in a statistically significant mannerBack to article page