Kirby continues to get basic maths wrong

23 Mar

As I’ve pointed out, Kirby has messed up his maths on the Huffington post blog. He’s now done much the same in an Atlanta Journal-Constitution article (clearly, their Editor does not check for mathematical errors, or general stupid). Kirby’s article is titled “Give us answers on vaccines”: I don’t expect any answers from Kirby, but correct sums would be a start.

Kirby argues that:

Most striking is how typical Hannah’s cellular dysfunction [mitochondrial disorder] may be among children with autism. While extremely rare in the general population, at two per 10,000 people, it seems unusually common in autism — with estimates up to 2,000 per 10,000.

To go over these figures again – a 2,000 per 10,000 incidence of mitochondrial disorder among people with ASDs is and incidence of 1/5. If 1/5 people with an ASD has mitochondrial disorder, and only 2/10,000 people have mitochondrial order, then if everyone with mitochondrial disorder has an ASD only 1/1,000 people would have an ASD. Clearly, not everyone with mitochondrial disorder has an ASD, so the actual incidence that Kirby is estimating would be considerably lower.

Age of Autism linked the blog post on Kirby’s article, and there are (at time of writing) over 170 comments. Oddly, though, it seems that no-one has picked up on Kirby’s basic error. Haven’t they noticed their hero’s issues with maths? Or don’t they care?


7 Responses to “Kirby continues to get basic maths wrong”

  1. Joseph March 23, 2008 at 03:50 #

    Who says 2,000 per 10,000 anyway? Can’t he say 20%? We went over that the other day. It is mathematically impossible for the prevalence of mitochondrial disorder to be 0.02% in the general population and simultaneously 20% in the autistic population. Most people should be able to see that. David Kirby is probably not interested in, you know, facts or arithmetic.

    Of course, it’s entirely possible mitochondrial disorder is under-recognized. We should know very well how that goes by now.

    Additionally, I think the problem is that he’s not really talking about mitochondrial disorder in autistic children, but about markers of mitochondrial dysfunction or something of the sort.

    In fact, David Kirby has not clearly stated where exactly he got the 20% figure, has he? I’d really like to know. The mistake will be obvious then.

  2. Ms. Clark March 23, 2008 at 09:41 #

    I think his excuse will be that they find mitochondrial disorders well enough in the general population, but if the mito problem sufferer happens to autistic, no one ever thinks to check for mitochondrial problems. …. because…. I guess because when autistic people have them there issues don’t look anything like mitochondrial prolems in normal people… or mito doctors hate autistic people and won’t diagnose them at all.

    I don’t think mito-anything is as common as 2% among all the autistics together.

  3. Jon March 23, 2008 at 20:32 #

    Sure – it’s quite possible that Kirby believes that mito disorders are underdiagnosed in autistic people. However, the way he writes makes it impossible to figure out what he thinks, and Kirby uncritically presents estimates of prevalence of mito disorders in autistic people and the general population even where it’s clear that they can’t both be right.

    And yes – presenting a percentage as 2,000/10,000 is a rather confusing way of doing it. But then all the stats here seems to be rather confused.

    I think I should bill Kirby for the head-shaped dent in my desk…

  4. Joseph March 23, 2008 at 20:56 #

    it’s quite possible that Kirby believes that mito disorders are underdiagnosed in autistic people

    What I meant is that mito disorders are probably underdiagnosed in general. So the prevalence is not really 0.02%.

  5. Prometheus March 24, 2008 at 18:43 #

    I believe that Kirby got the “20%” number from a quote by a Dr. Kelley, who claimed (without providing data) in an interview that “Mitochondrial disease was the most common diagnosis, accounting for as many as 20% of autistic children.”

    Now, it is but a short step on the calculator to go from an off-the-cuff comment about “…as many as 20%…to “…estimates up to 2,000 in 10,000…”.

    What Kirby doesn’t say is that there is – so far as I know – only one estimate that goes that high. Most of the other estimates are far lower. A more accurate way to phrase this “factoid” would be:

    “While extremely rare in the general population, at two per 10,000 people, one doctor estimates it may be as high as 2,000 per 10,000.”

    That’s more accurate, if less exciting.

    Of course, if Kirby had taken any science classes in college, he would have known that this is a silly way to represent such a high prevalence. He also would have known that one doctor saying “20%” to a reporter is not a reliable number. Data, please.

    Still, you’d think he would have at least “run the numbers” once he had turned on his calculator. After all, if “20%” of autistic children have “mitochondrial disease”, that would represent a 1000-fold increase in risk. With all of the genetic research going on in autism, I would have expected someone to notice a huge linkage between the mitochondrial genes and autism, as oppsed to the small and variable linkage currently known.

    I’d also be interested to know what percentage of people without autism Dr. Kelley finds have “mitochondrial disease.” The “mitochondrial disease” subgroup of “alternative medicine” seems to find a lot more “mitochondrial disease” than any researcher in the field. Perhaps that’s due to the tests they use.


  6. David N. Andrews M. Ed. (Distinction) March 25, 2008 at 09:03 #

    “And yes – presenting a percentage as 2,000/10,000 is a rather confusing way of doing it. But then all the stats here seems to be rather confused.”

    Actually, the n/10,000 format is pretty standard in epidemiology work, but that’s because the values of n tend to be in single or double figures. When it gets beyond 99/10,000, a better (or, at least, a less confusing) way to do things is to count per mille or per cent (n/1000 or n/100).

  7. Jon March 25, 2008 at 11:31 #

    Ah – thanks. I guess 2/10,000 is a sensible enough way to express the figure, but 2,000/10,000 does seem odd…

Comments are closed.

%d bloggers like this: