The other big component is using Bayesian inference. For that you need a prior. What's the prior for this? You could look at incidents the US has had in simi

The other big component is using Bayesian inference. For that you need a prior. What's the prior for this? You could look at incidents the US has had in similar freedom of navigation actions, what China has done with other countries, or other similar situations. For example, I can only think of a few similar incidents, and none that escalated to fatalities at sea. So your prior would probably be pretty small, say 1%. After that you would adjust according to new data. For example, China seems willing to do more risky stuff lately, with India being a perfect example. You might raise your estimate based on that. So maybe that raises your estimate to 2%. But it probably shouldn't completely alter your estimate.