Thread by @sophie_e_hill

Harvard will continue employing David Kane, who has never published in a statistics journal and has a history of making basic stats errors, as Preceptor of Statistical Methods.

This isn't just failing upwards. This is affirmative action for racists.

thecrimson.com/article/2020/9…

The 🧾's:

In 2006, David Kane wrote a blog post on the Harvard @IQSS website, accusing the authors of a study published in Lancet of fraud.

The Lancet study, which estimated excess deaths in Iraq due to the war, involved a survey with a near 100% response rate.

Kane argued that this was indicative of fraud.

(Note: Kane's blog post is no longer available but it was quoted on @crookedtimber)
crookedtimber.org/2006/10/18/flo…

Now, to be clear: fraud by survey enumerators -- especially in low-income countries -- is a real problem!

But to detect it we need actual evidence - like examining the proportion of near-duplicates in the survey (see Kuriakose & Robbins, 2016).

mdhrobbins.net/uploads/6/2/8/…

Kane's only piece of evidence for saying the study was fraudulent was the near-perfect response rate.

But, as many commenters pointed out, response rates vary based on context and several other surveys in Iraq also had near-perfect response rates.

crookedtimber.org/2006/10/18/flo…

As @kjhealy points out, it's hard to understand why Harvard @IQSS would allow Kane to use their blog to make an evidence-free accusation of fraud about a peer-reviewed study.

crookedtimber.org/2006/10/18/flo…

Harvard @IQSS responded by immediately taking down the post and apologizing, saying that "tone is unacceptable, the facts are shoddy, and the ideas are not endorsed by myself, the other authors on the sidebar, or the Harvard IQSS".

crookedtimber.org/2006/10/18/fra…

So, what happened next? Did Kane apologize for making the accusation? Did Harvard @IQSS reconsider its relationship with Kane?

Nope! In fact, Kane *doubled down* and wrote an entire paper critiquing the Lancet study. He remained a fellow of @IQSS until 2012.

After giving up on the accusation of fraud, Kane tried another tack.

He wrote a paper (with the help of 4 research assistants!) arguing that the Lancet study could not rule out the possibility that excess deaths in Iraq went *down* due to the invasion.

web.archive.org/web/2011101502…

How did Kane reach that conclusion?

Step 1. He notes that the Lancet study includes data from Fallujah, which is a clear outlier due to the extensive bombing campaign there

Step 2. Argue that once Fallujah is included, the variance of the estimate of post-invasion mortality has increased so much that the confidence interval for difference (post-invasion mortality - pre-invasion mortality) must cover 0.

This is... quite a silly argument. Kane provides two mathematical "proofs".

"Proof" #1 assumes that pre- and post- mortality rates are normally distributed.

"Proof" #2 generalizes this by assuming that pre- and post-mortality rates are unimodal.

There's just one problem... based on both the raw data and our understanding of the data-generating process, we *know* that the distribution is neither normal nor unimodal!

Kane *almost* acknowledges this point when he notes that assuming a normal distribution is technically incorrect since mortality can't be negative.

But the concern is quickly dismissed, since Kane tells us that the calculations are similar with a truncated normal distribution🤦

This is like saying that if you do a survey of the annual earnings of Harvard dropouts and one year you happen to sample Mark Zuckerberg, then that gives you reason to think that the annual earnings of Harvard dropouts might have *decreased*.

Kane repeatedly tried to argue that he was only following the assumptions made in the Lancet study, and thus his critique was really one about the internal consistency of the paper rather than the appropriateness of any assumptions.

Unfortunately this just isn't true. The authors of that study did not make these assumptions, as this commenter pointed out (at 3:48AM... not all heroes wear capes).

So David Kane wrote a paper accusing the authors of a published study of making very silly mistakes. Commenters then pointed out that it was in fact *he*, David Kane, who was making the very silly mistakes.

So, did he apologize and retract the paper?

Nope. Instead, he gave permission to Michelle Malkin to post it on her website, where the idea that "we cannot reject the null hypothesis of no increase in excess deaths in Iraq" reached a mass audience.

unz.com/author/michell…

Note: Michelle Malkin is a white supremacist.

(Gosh, David Kane does seem to accidentally bump into contact with well-known white supremacists quite a lot, doesn't he?)

https://twitter.com/sophie_e_hill/status/1310623038585663494?s=20

This whole kerfuffle happened in 2006/07, when Kane was affiliated with Harvard @IQSS. He remained a Fellow there until 2012, and was later hired in 2018 by the Harvard Government department as a Preceptor.

What does this man have to do in order to *not* be taken seriously??

This one is too funny not to mention.

In 2010, David Kane was banned from editing articles on the topic of "race and intelligence" on Wikipedia: en.wikipedia.org/wiki/Wikipedia…

I thought that perhaps I had caricatured David Kane when I described him as a racist middle-aged man who is creepily obsessed with his undergraduate alma mater.

But judging by his Wikipedia edit history... damn, I'm good.