Is vaccine efficacy a statistical illusion?

Just a twitter-ish one liner:

Is vaccine efficacy a statistical illusion?

Insight delivered on a plate. A clear explanation of the John Dee's Almanac concept. Look how the sizes of populations shift with time on a fixed death rate giving the illusion of efficacy. And also of apparent waning efficacy with time. So elegant, so neat, love it.

Peter

Addendum if it helps:

EDIT Just to clarify, there is no need for the "vaccine" to do anything, you can even assume it's a placebo injection. The effect still occurs. END EDIT

After a chat with Raphi on twitter this might make it clearer. Campaign starts at day one. No results are collected for a week 'cos that's how long it takes. No one know exactly when a given person died because mortality stats are like that and this is not a controlled study situation we're talking about.

The numbers of deaths collected a week after the campaign started are attributed to week two because that's when they are recorded. This is the source of the error.

If 15 people a day die during week one but are recorded as week two they will be put in to incorrect population sizes because the vaccinated population is rising rapidly and the unvaccinated population size is falling rapidly. A week is a long time in a vaccine roll out.

So the small number of deaths in the initially tiny vaccinated group of week one will be attributed to the significantly larger vaccinated group found in week two. Very few deaths from a very small population are now spread out over a now larger population.

The much larger number of deaths from the much bigger unvaccinated population of week one will be attributed to the now smaller unvaccinated population of week two. The population is smaller because vaccines have been given, which rapidly reduces the size of the unvaccinated population.

In the vaccinated group too small a number of deaths is spread through too large a number of people, hence a low incidence/person days. Vaccine appears to work.

In the unvaccinated too many deaths (it was a very big group in week one) are attributed to a population reduced by the number who have been vaccinated by the rollout. So a much higher figure per person days is found.

Don't start me on how this makes being unvaccinated intrinsically dangerous and how the 'rona vacc appears to protect agains all cause mortality. Just more artifact.

The graphs come out as in the linked blog post.

The need for graphing mortality curves by date of death vs date of reporting is well known from plotting peaks of waves from peaks of deaths. If a study uses date reported rather than date of occurrence, it's possibly junk. It can take months to get death numbers by date of occurrence vs reported in the real world. Some mortality data from the UK ONS will be delayed by the time needed for a coroner's inquest.

Peter