I am sorry that when i posted this before, it was attributed to another author. We are sorting out how this has been done
Dear friends,
I have been digging a little deeper concerning estimating deaths during disasters such as with Maria. One thing I am very much struck, is how variable the estimates are from developing countries. There was an enormous amount of variability of estimate for the Tsunami and Pakistan earth quake. The problem involves not only counting deaths, but many of the countries did not have adequate census information.
A expert disaster friend of mine indicated that there was a new report from the National Academies of Sciences describing assessments of populations at risk, risk factors, deaths, injuries and costs. It is interesting that they did not discuss the indirect approach presented by Harvard and GW.
https://www.nap.edu/catalog/11895/tools-and-methods-for-estimating-populations-at-risk-from-natural-disasters-and-complex-humanitarian-crises
The BMJ just published an article on monitoring death tolls in disasters
https://www.bmj.com/content/362/bmj.k4005.full?ijkey=MxStM9ncBv27vyW&keytype=ref
Finally, I contacted a good friend of mine, George Seber. Many of you know George as the guru of measuring animal abundance. I feel in love with Capture – recapture.
"Hi Ron, I am not into stats right now but writing about alcohol. Hence a top of the head response.
The second study used a random selection of households from a number of strata. I think stratification should perhaps be tied to the path of the hurricane, i.e. according to areas likely to have various stages of damage (perhaps concentric regions, though this might be hard to carry out). Then in each strata I would recommend adaptive sampling starting with an initial random sample of households. A neighborhood of a household is the defined (a group of households). If there are death at the chosen household, then the neighborhood is searched for further deaths, and if a mortality is found we then search the neighborhood of that household. We then end up with a cluster of households in which there are deaths surrounded by households with no death. The first report suggested other nonstatistical problems. George Seber"
I have come to the conclusion that we are just not very good at obtaining estimates of mortality. However, interesting, once a country says there are 187,321 deaths, this is a figure reported over and over. The Tsunami is likely the poorest count as so many people were swept to sea, and the quality of the census data was so variable, and actually changed quite a bit.
I am still suspicious of the indirect measures as they are unreasonably high. The query of the Undertakers was quite interesting.
What I have see to get better figures we need to have disaster experts, statistical and epidemiologic to meet to discuss this. Perhaps a session at the ASA meetings would be appropriate
------------------------------
Ronald LaPorte, Ph.D.
Professor Emeritus Epidemiology
University of Pittsburgh
ronaldlaporte@gmail.com724 934 9023
------------------------------