My recent post, What do the anti-vaxers want, generated a lot of commentary from the anti-vax crowd, much more that I had expected to be honest. I do not intend to go over that commentary here. What I intend to do is to analyse a study that was linked to by one of the anti-vax enthusiasts. This study has been making the rounds lately in the anti-vax crowd (hello Mike “The Health Danger” Adams) because it purports to show a relationship between vaccine doses received and Infant Mortality Rates (IMR). The study in question has been published here, and David Gorski has done a thorough review of it over at Science-Based Medicine blog. I will not repeat what he wrote.
What I intend to do here is to show how easily numbers can be manipulated to show a trend which can then be touted as proof of a relationship. We always say that correlation does not equal causation, yet that is hard for some people to grasp without an example. This study provided the perfect example to illustrate the concept. So without further ado, let us show why this study does not really show anything in connection with vaccines.
What does the study say?You can read the full thing yourself, but in a nutshell the authors looked at IMR in the U.S. and noticed that for the year 2009 it ranked 34th. They then took the IMRs for the other 33 countries that did better than the U.S. Then they counted the vaccine doses recommended for each country (they used a funny way of counting DTaP as 3 doses each time, but we’ll let that slide for the sake of our exercise). Then they plotted, in a graph, the number of vaccine doses vs. the IMR rate for each country. The results were SHOCKING:
Sunday, May 15, 2011
Recently I reposted the What do anti-vaxers want article from Vaccine Times. Well, he has a follow up on that. I think this plays well into the famous Mark Twain quote about "Lies, damn lies, and statistics" when dealing with data. That is not to say that data is inherently wrong. It's more about how people will torture it to come up with a preconceived notion as opposed to letting that actual data lead you to a discovery about the behaviour of reality. I think Leart does a great job on this post.