Monday, September 13, 2021

A question from a statistics amatuer

So the news exploded today that the unvaccinated are 11 x more likely to die of Covid than the vaccinated.  It was met with multiple stories of medical pros beseeching everyone to become vaccinated.

Yet I remember a few months ago, and they said that between 97% and 98% of all who died of Covid were unvaccinated.  

Now I can do arithmetic in my head as well as anyone.  I'm actually quite good at it.  But my math and stats skills drop from there.  To my amateur eyes, however, that seems to be a big difference.  It seems as if the effectiveness has dropped. 

From my vantage point, the earlier stats that would be like saying only about 1 or 2 out of every 100 people who die of Covid are vaccinated.  But now, for every 11 unvaccinated who die, there is 1 vaccinated who dies. Which is a lot more than the old 1 or 2% who are vaccinated who are dying.  It's more like around 9 or 10 vaccinated who die for every 100.  Which, if I'm not missing something, is significant. 

Am I reading this right?  Am I reading the numbers correctly?  Are they presenting them in such a way that isn't the best way and that's why I'm confused?  Or am I just a Charlie Brown when it comes to this? Any help from numbers-pros is appreciated. 

7 comments:

  1. Don't trust any statistics that doesn't give you the uncertainty associated with the conclusion. Further, don't trust multivariate statistics (e.g. people) unless there's lots of caveats about missing variables (or it's physics). Your math is pretty good, but it's sufficiently gross that you can't say much other than there's a difference between people dying then versus now. The medical field tends to fail to control for all selection biases.

    ReplyDelete
    Replies
    1. That seems about right. I get the feeling we're given a molehill of info and a mountain range of info is ignored. I asked my wife about this, and she said one problem is they meander around, picking one stat here, and then picking another there, even when there is no real connection. Sort of comparing apples and carburetors.

      Delete
  2. The problem is not really with statistics. You quote no one reporting a p-value to determine how likely the result is to be from chance, or performing a Bayesian analysis, or doing any of the other things real statisticians would do. This is more about data collection, especially when polls are taken. Many polls today are not taken to collect data, but to advocate a position, either directly to the person being polled or to those who are subject to peer pressure from polls.

    A bigger problem is how statistics are reported in the media. Reporters not only have agendas, they also rarely understand anything technical or mathematical. When they start paraphrasing a study, or telling us what the study "really means", it no longer really matters how good careful the research was.

    And in spite of all the bellyaching, most of the people who heap disdain on statistics still use it when they think it supports their position -- for instance, showing that a majority of Americans oppose abortion past a certain point of development, or that Biden's popularity has tumbled as the situation in Afghanistan deteriorated, or that police are not really targeting blacks with deadly force to an extraordinary extent.

    ReplyDelete
    Replies
    1. "it no longer really matters how careful the research was"

      My proofreading was not good careful.

      Delete
  3. My understanding is that the percentage quoted of "those who died that were unvaccinated" includes 100% of deaths with COVID, from the beginning of the pandemic. If you want some estimate of vaccine effectiveness, look for something like "percentage of Israelis who died in the last four weeks with/of COVID. Israel is one of the most vaccinated nations on the planet.

    ReplyDelete
  4. It is two different statistics. The 97% number is percentage of those with Covid who have died whether vaccinated or not. The other is the relative percentage.

    So lets say that only 20% of people are vaccinated. There are 5000 people, so 4000 unvaccinated and 1000 vaccinated. The vaccinated have a .1% chance of dying, so one person dies. The unvaccinated have 11 times the rate of dying, so 1.1%, out of 4000 people means 44 people. Then the percentage of those who died while unvaccinated would be 44/45 or 97.8%. So the two numbers could be true at the same time.

    ReplyDelete
  5. Thanks all! It looks as though there is nothing particularly wrong with any of the stats, but there's clearly a maneuvering around of different approaches to the stats, and likely not by accident.

    ReplyDelete

Let me know your thoughts