I've revised my Infection Fatality Rates (IFR) from the Spanish Antibody Study to make it more accurate

Main changes:
1) data from different dates to correspond with when samples were taken
2) added confirmed cases to the denominator
3) added probable deaths into the final IFRs

Change 1: New data period to coincide with actual dates the serological samples were taken. Samples were taken over a 15 day period between April 27 and May 11.

Based on this chart I chose 18 days as an offset where positive rate for IgG is close to 100%.

Change 1 (cont): Because samples were gathered across 15 days, I chose a period offset by 18 days but also covering a 15 day period.

Therefore I took the data for all 15 days between April 9 (18 Days before April 27) and April 23 (18 days before May 5) and created an average.

Change 1 (cont): Is it correct to use 18 days? A little more than half will have antibodies by the end of one week.

BUT, keep in mind the chart is from symptom onset and "Symptoms may appear 2-14 days after exposure" https://www.cdc.gov/coronavirus/2019-ncov/symptoms-testing/symptoms.html

Change 2: I added confirmed cases to the total infected population predicted by AB testing.

So projected number of infected as per AB testing + confirmed cases is now the new Denominator

Change 3: I added the probable deaths into the total deaths that make up the Numerator.

Over the 15 day period, and also thereafter there is a pretty consistent number of probable deaths so the number isn't growing.

Change 3 (cont): Last day this data is available (5/18) there were 8523 probable deaths.

The average for the 15 days was 8678.

But on May 25th they revised deaths down by 1915 so I subtracted this from the total probable deaths bc unconfirmed deaths is where they would cut

Change 3 (cont): So that left me with 6763 probable deaths which I redistributed proportionally based on the percentage of total deaths each group represented.

Finally, here are the sources:

Spain AB Study: https://elpais.com/sociedad/2020-05-13/solo-un-5-de-los-espanoles-se-han-contagiado-de-coronavirus-segun-los-primeros-datos-del-estudio-de-prevalencia.html

Population: https://worldpopulationreview.com/countries/spain-population/

Case Data:
https://www.mscbs.gob.es/profesionales/saludPublica/ccayes/alertasActual/nCov-China/documentos/Actualizacion_71_COVID-19.pdf other 14, change the 71 +1 to get next day

Antibody prevalence data: https://www.nature.com/articles/s41591-020-0897-1

Original chart https://twitter.com/gummibear737/status/1262453620735385605?s=20