Monday, April 19, 2010

The Anti-Vax Movement Still Peaked in 2002-2003 and MJ's Excuses are Trivial to Address

MJ has written yet another rebuttal of my post on the media's interest in the anti-vax movement.

MJ's primary argument is basically that the absolute article count for "autism vaccines" has grown. The count for "autism" has simply grown more. I say that the relative count is what matters, but let's look into this in more detail.

As I've noted in comments, MJ fails to take into account that the total number of articles indexed by Google News Archive has also grown from year to year. Presumably, this doesn't mean people are reading more newspapers, but simply that Google is adding sources to its index all the time.

Estimating the total number of articles in Google News Archive could presumably be problematic algorithmically for Google. I have some ideas which I mentioned in comments, but I'm not confident that they would be unbiased.

Instead, let's check if searches for "autism genetic" relative to "autism" have the issue MJ thinks relative counts have.



They do not. There's no peak in 2002 here. There's no downward trend after 2002. There's no dilution of the word "genetic" in autism articles as more topics are covered. What we see instead is a remarkably stable trend.

We previously also looked at relative counts for "neurodiversity", which MJ had verified as well. This analysis also fails to support MJ's hypothesis. The same is the case of several other trends I've checked which I'm not going to go into here.

It's also illustrative to look at the raw article counts for "vaccine injury" (in quotes). These counts are presumably also biased by increasing coverage of autism topics, but maybe less so.



What we see here is consistent with a peak in 2002-2005, and a brief recovery in 2008 due to substantial propaganda efforts in relation to Jenny McCarthy and Hannah Poling. This effect cannot be expected to last very long, though.

MJ has failed to explain why the pattern of media coverage for "autism vaccines" generally matches VAERS autism submission trends and the number of autism cases filed with the vaccine court. It's lazy and convenient to simply say "the data was not meant to track this." What is the explanation?

Additionally, Smith et al. (2007) reports that:
MMR vaccine remains the number one ‘top of mind’ vaccination issue for parents. The proportion of parents believing MMR to be a greater risk than the diseases it protects against has fallen from 24% in 2002 to 14% in 2006. The proportion of ‘hard-core rejectors’ of MMR vaccine remains stable at 6%. There has been a gradual and sustained increase in the proportion of parents across all social groups saying MMR was completely safe/slight risk rising from 60% in 2002 to a current level of 74%. There now appears to be a sustained move away from fears over MMR safety and belief in the unfounded link to autism towards a more positive perception of the vaccine.

(My emphasis.)

Finally, an immunization report by the British NHS shows that MMR coverage had a low in the 2003-2004 period.

Conclusion

I've perseverated on this a lot more than I probably should, and I've gone out of my way to address ridiculous criticisms put forth by MJ.

12 comments:

  1. You still don't get the point that you are effectively measuring market share instead of interest. So, lets try this as an example.

    If we are looking to see if sales of a particular car are increasing, I would look at the sales figures for that model of car. You on the other hand would look at the market share (percent of total car sales) the car had. But a car's sales could still be going up and at the same time its market share could be going down.

    I would say that the sales are still growing, just not as fast as the competition. You would say that the sales of the car are shrinking.

    But you certainly can't say that a car with a 20% market share one year and a 10% the next had decreased interest if sales were actually up 15% over the same time period.

    The same thing is happening with autism, the market (news articles) are growing faster but this specific part (vaccine/autism), but that doesn't mean this part is shrinking.

    You can say that your real interest is the market share of the idea, but that wasn't what your initial post was about.

    BTW, you are not quite correct about what google news archive does with the AND. Try searching for 'autism and vaccines' (13,000) and compare that to 'autism "and" vaccine' (10,700). The second form treats AND as a literal but the first term uses it to restrict the results. In the case of 'autism vaccine' and 'autism and vaccine', it doesn't make a difference, but in others searches it does. Google used to completely ignore AND but I think they changed is somewhere along the lines.

    Regardless, using AND would underestimate the number of articles rather than skew it "in my favor", as you suggested.

    ReplyDelete
  2. You still don't get the point that you are effectively measuring market share instead of interest.

    And you still haven't addressed why that would work fine for "autism genetic" but not "autism vaccines." It could be anything else, like "autism boys." Why should we give special consideration to vaccines?

    I would say that the sales are still growing, just not as fast as the competition. You would say that the sales of the car are shrinking.

    I would say that people's interest in that particular car is diminishing. They are becoming more interested in other cars. Whether the sales volume of that particular car is still increasing simply because more cars are being sold in general is not what I'm interested in. But also, you haven't even shown that sales of that car are actually increasing, because you have not controlled for inflation.

    BTW, you are not quite correct about what google news archive does with the AND.

    Not important, but I don't think so. Try searching only "and".

    ReplyDelete
  3. "And you still haven't addressed why that would work fine for "autism genetic" "

    What do you think the relationship between these terms is? Just because one grows faster than the other does not mean that both aren't growing

    I am positive that there are other terms that are growing faster than the relationship between "autism genetic" and autism - there have to be. Either that or there are a large number of new terms entering the autism superset. After all, the "autism" search is the superset and each of the others is a subset and if you add up all of the subsets, you get the entire set. If one subset isn't growing as fast as the entire set then something else is filling the void.

    Or are you taking the absurd position that the media interest in autism hasn't skyrocketed in the past 13 years?

    "I would say that people's interest in that particular car is diminishing. "

    And you would be wrong. The fact that interest in other cars might be increasing faster than this model does not mean that interest in this model is shrinking - unless you are talking about market share.

    Think about finance, if a stock under performs its market, that doesn't mean it hasn't grown, just that it hasn't grown as much.

    ReplyDelete
  4. Let's try a more specific example. Toyota is losing trust, right? Suppose there's a sudden boost to the economy, and people start buying cars left and right, and as a result, even sales for Toyota increase. Would that mean people are becoming more interested in buying Toyota?

    ReplyDelete
  5. That is exactly why you look not only at the raw numbers but also at how the raw numbers are changing (the delta). This is where concepts like YoY changes and moving averages come in. Tools like this help you see the overall trend in the numbers more than concepts like market share do.

    So, in your example, if the sales grew but grew less than would be expected or you could point to a decrease in growth you could say that interest is slowing. But even then you wouldn't say that the overall trend has reversed and people are no longer interested - just that interest is decreasing.

    Do yourself a favor and pick up a good book on technical analysis, these books have chapters and chapters devoted to trend analysis.

    But you still haven't addressed the issue of relative vs absolute growth. Are you talking about the market share of the vaccine/autism idea or the media's overall interest? Nor have you said whether you think the overall autism numbers are heading up or staying the same.

    ReplyDelete
  6. This is where concepts like YoY changes and moving averages come in. Tools like this help you see the overall trend in the numbers more than concepts like market share do.

    MJ, you talk as if I'm not familiar with moving averages or time series analysis, and that might be because I haven't used them in this blog, but I am. There are more sophisticated means of smoothing, e.g. a spline interpolation.

    I frankly think it's nonsense to try to apply a moving average to 13 years of data that is not very noisy. What's the point?

    If you really want to determine if there's a trend, you need to calculate the slope of the series and its confidence interval. You could even go further and control for auto-correlation, but that would be ridiculous for such a simple exercise.

    That aside, the example stands. People's interest in Toyota and Toyota sales are different things. The way you'd measure it is exactly by looking at whether Toyota is losing market share.

    But as I noted, you're also taking advantage of the fact that the number of articles indexed by Google grows from one year to the next.

    But even then you wouldn't say that the overall trend has reversed and people are no longer interested - just that interest is decreasing.

    And how is this inconsistent with what I said?

    ReplyDelete
  7. Hi Joesph - I'm Rebecca from NationalAutismResources.com and I stumbled across your blog the other day as I was searching for sites to feature on own blog at, http://www.NationalAutismResourcesBlog.com/ We think your blog fits into what our customers and readers would be interested in. If you are ever interested in exchange blog posts about one another's sites, please feel free to email at, nationalautismresources1@gmail.com
    Thanks for your time,
    Rebecca, NationalAutismResources.com
    http://www.NationalAutismResources.com
    http://www.NationalAutismResourcesBlog.com/

    ReplyDelete
  8. Joseph,

    too bad the EOHarm group doesn't go back far enough to give you another measure.

    That said, it does go back to 2005. If one looks at the traffic there (I believe you have used this metric before), it is clear that the interest level is down. There was a peak in 2007 with about 4000 messages in a month. Traffic is down to about 1/10th that peak level.

    ReplyDelete
  9. @Sullivan: Yeah, I've noticed that. But then again, maybe a lot of the EOHarm regulars have moved to AoA. Smith et al. (2007) indicates hard-core rejection of vaccines remains stable.

    ReplyDelete
  10. Hey Joe,
    Is there a way I can get in touch with you outside of this blog?
    Thanks
    Phil Commander

    ReplyDelete
  11. Joseph,
    This may be on topic, or maybe slightly off, but I haven't seen you tackle the correlation between autism and larger than normal head circumference in infants. I am curious to hear your opinion on this recent finding.
    I'm sure you are aware of it, but in case you may have missed it here is a link:

    http://www.sciencedaily.com/releases/2008/01/080130140127.htm

    Thank you.

    ReplyDelete
  12. Yesterday, the paper called Functional Impact Of Global Rare Copy Number Variation In Autism Spectrum Disorders came out in Nature http://www.nature.com/nature/journal/vaop/ncurrent/full/nature09146.html
    Can you, with your experience and background, comment on the paper in light of your hypothesis that autism is merely and expression of Natural Variation and not a disease per se. Your fan, Julia

    ReplyDelete