• Melanie wrote a new post, Bad Chart Thursday: New Years Linked to Autism, on the site Skepchick 5 years, 4 months ago

    ThumbnailShocking news from the unironically named website The Mind Unleashed: “MIT Researcher’s New Warning: At Today’s Rate, Half Of All U.S. Children Will Be Autistic By 2025.”*

    Autism: a fate worse than death for […]

    • Bonus benefit of getting rid of your calendars and clocks: you won’t age!

    • “Is it the use of two y-axes whose values can be adjusted to make the data fit as closely as you like?” Just elaborating on this point a bit, both overlaid graphs show approximately exponential growth. Lots of things show approximately exponential growth. E.g., Moore’s Law describes the change in processing power of computer chips as exponential growth. And, when rescaling the y-axis, all exponential growth curves can be made to look the same.* So I could construct a similar graph to show that the rise in autism follows a similar curve to the rise in computer processing power, and by similar reasoning argue that faster computer chips are causing autism. This point should be extremely obvious to any research scientist, so creating a graph like this one displays a willingness to use numbers to distort rather than inform. It is not inconceivable that chronic low doses of environmental toxins like pesticides could contribute to autism, but this is not evidence for that, and I’m not aware of any such evidence that is actually based in solid research rather than irresponsible speculation.

      * Explanation: the general exponential growth formula is y = b * e^t, where y is the response variable, t is the independent variable (usually time), and b is a growth constant. Rescaling the y axis of any graph is equivalent to replacing y with c*y, where c is the amount by which to rescale. But this gives c*y = b * e^t, or y = (b/c) * e^t. Since b and c are both constants, this is the same as the original growth formula with a new value for b. So if I have two approximately exponentially growing phenomena with growth constants b1 and b2, I can always make them look the same by rescaling the graph of the second one by c such that c = b1/b2.

      (BTW, sorry if there’s a double post on this — the site is really slow and from Cloudflare it looks like Skepchick might be getting DDOSed, so I’m not sure if my first post took.)