Skip to content

Driving a stake through that ages-ending-in-9 paper

David Richter writes:

Here’s a letter to the editor [in PPNAS] in response to the ‘people with ages ending in 9’ paper?

We point out some problems with their analyses and their data and tried to replicate their theory in a large German panel study using a within-subjects design and variables close to those used in their paper.

We found no evidence for their theory and this is perfectly in line with your blog post from today [18 Sep 2016]: if hypotheses were ‘true’, they should be ‘true’ independent of the data sources used.

My reply: What’s particularly ridiculous about that ages-ending-in-9 paper was that even their own data did not particularly support their hypothesis (as I discussed in that post from a few years ago).

It’s a sad reflection on the state of the American science establishment that this sort of obviously bad work (the original ages-ending-in-9 paper, that is) was published by the National Academy of Sciences.

P.S. The title of this post is a nod to Jeremy Freese’s description of certain claims as “more vampirical than empirical: unable to be killed by mere evidence . . . the hypothesis seems so logically compelling that it becomes easy to presume that it must be true, and to presume that the natural science literature on the hypothesis is an unproblematic avalanche of supporting findings.” Also the idea that any effect could go in either direction and support the story. Suicide rates go up? That’s a sign of the despair of impending mortality. Suicide rates go down? No problem, it’s a sign that people are valuing what they have. And so on.


  1. Anoneuoid says:

    I’m not sure what this means: “replicate their theory”. Wouldn’t you usually replicate results/observations but test a theory? Perhaps some kind of translation issue?

  2. Erik says:

    I submitted my criticism on their own data as a letter to the editor (at PNAS) back in 2014, but they didn’t find it suitable for publication. It got published as a comment in Frontiers in Psychology instead:

  3. Marcus says:

    Andrew, you obviously forget that Germans are entirely unaffected by the ages-ending-in-nine effect ;-). This is simply a cultural/linguistic moderator that we simply need to explore further – preferably with a very large NSF grant.

  4. Jonathan says:

    If they did this in China, would there be more major life changes at a 4 age? Wouldn’t 9 then signify longevity, etc. so stay the course? Maybe 4 would mean don’t do anything because you might be unlucky in a 4 year, especially if your birthdate ends in 4 and you’re a 4 age. The complexity of pathways, their number and the way they spread through various chains of possible causation, means you can find some effect or not. But it is kind of cool to think that you can find an effect no matter which direction you set the sign! Gives another depth to the concept of ‘relative’, don’t it?

  5. Ben Prytherch says:

    I dunno, check out their p-values:

    0.632, 0.109, 0.131, 0.112, 0.206, 0.349, 0.850

    If there were no age effect, these should follow a uniform(0,1) distribution, which has a mean of 0.5. But the mean of these p-values is 0.34, and significantly less than 0.5 (t(6) = 3.09, p = 0.02). Looks like there’s a 98% chance that the “ages ending in -9” hypothesis is true after all!

    • a reader says:

      Actually, you’ve shown that their p-values are significantly greater than 0. Testing whether the mean of the p-values is less than 0.5 results in a p-value of 0.2008. Hence the null that they are all null must be true.

      • Ben Prytherch says:

        Hi a reader, I appreciate the interest you’ve shown in my work. Science advances only when we engage in robust dialog, and I welcome the opportunity to further explore the implications of my original analysis.

        The results I reported should have read (t(6) = -1.44, p = 0.2). This does not change the conclusion of the analysis.

    • Jonathan (another one) says:

      And the Kolmogorov-Smirnov test against a uniform(0,1) has a p value of 0.067. Not even close to 0.05.

      • Andrew says:


        Add the digits of this p-value and you get 13. Yes, it’s unlikely that the sum of the digits of the p-value, taken to three significant figures, will be prime—but the result is not statistically significant at the 5% level. Therefore the null hypothesis is true, and the planned Ted talk will have to be canceled until further notice.

        • Dan Simpson says:

          You’re peddling disinformation here Andrew. Once the blood ritual is complete, you cannot cancel the TED talk. If you do, Ted Turner, for whom the talks were named, stands in Central Park and loudly summons Baphomet to destroy what’s left of your soul. Similarly, when people don’t retract TED talks it’s really because they’re trading off a lifetime of unscientific infamy against an eternity of fiery damnation.

Leave a Reply