Figure 2

Bayesian dependency network of 10 characteristics in 520 severe asthmatic patients. An edge between two nodes (i.e. characteristics) represents direct casual relevance, and its width is proportional to the probability of the corresponding nodes being directly relevant to each other (i.e. taking into account both possible edge directions assuming an underlying Bayesian network). Those connections are depicted where the a posteriori values between characteristics are larger than 0.15. The exact values can also be seen in Table 2. Positive (+) sign on an edge means that the direction of the relation is positive, e.g. patients with higher “best FEV1” have higher chance of having better “worst FEV1” value. Negative (−) sign indicates that the direction of the relation is opposite, “A” indicates adult, “M” male. E.g. “A-“ on the edge between “Disease onset” and “Allergy” means that allergy associates with childhood onset and not with adult onset. These relations can be seen in Fig. 3 and Table 3.