A quick followup to Rafa’s analysis of the death toll from Hurricane Maria, from Axios:
Puerto Rican Governor Ricardo Rosselló ordered a recount Monday of every death on the island since Hurricane Maria made landfall on September 20, as evidence continues to show that the official death toll grossly undercuts the true number, reports the New York Times.
There are at least two ways to do this. One way is inferential in nature, taking a look at what we might expect the mortality to be and looking at what was observed. This is tricky due to a variety of potential confounding factors and also comes with it statistical uncertainty. It seems the Governer is going to take a more “census-like” approach:
The recount will require interviewing doctors and family members of the dead to learn whether their cause of death could have been linked to the fallout from the storm. For example, a heart attack may have been brought on by the stress of the hurricane, or roads leading to the hospital may have been blocked by debris.
This approach does not have statistical uncertainty but has uncertainty of a different kind—separating proximate cause of death from ultimate cause of death. It’s going to be difficult either way.