The reasoning described thus far is sufficient to determine hypotheses that are temporally consistent. The next step is to compute the probabilities of the hypotheses. This is complicated because the hypotheses are actually constrained patterns of possible scenarios. The different scenarios within each pattern will have different probabilities. The issue is how precisely a hypothesis needs to be defined. For example, the pneumonia hypothesis does not specify whether the LVF was ever low and neither high LAP hypothesis specifies whether the PC came on immediately or after an hour or two. The first of these situations results in different nodes in the hypothesis. The second has different time intervals on the nodes. There is enough knowledge in the model to distinguish between either of these situations, but there is no reason to distinguish beyond what is clinically relevant. In the following we will require the nodes to be fully specified, but not the time intervals. Thus the problem is to compute the maximum probability represented by a path through the nodes of a hypothesis.
In all of the hypotheses, MI and nitroglycerin are true and not dependent on any other nodes. Their probabilities will be ignored in the following analysis since the probabilities are only used to rank order the hypotheses. We will first consider the two hypotheses with high LAP.