Posts from June 2020 (17)

Team Ratings for Round 1

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Predictions for Round 1

Here are the predictions for Round 1. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Team Ratings for Round 5

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Performance So Far

So far there have been 32 matches played, 20 of which were correctly predicted, a success rate of 62.5%.
Here are the predictions for last week’s games.

Predictions for Round 5

Here are the predictions for Round 5. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

I saw this ‘infographic’ on the Twitter feed of a civil rights lawyer in the US

The implication is so obviously wrong that you’d wonder how someone would fail to notice: 58% is about three times 19%, not seven times.  I wondered if this was a confusion of rate ratios and odds ratios, but that didn’t explain it either. So I went to the New York Times story, and saw this at the top of the page

Same numbers in the graph: a three-fold difference. Almost the same claim in the text —  a seven-fold difference — but now with a reference group.

If you read carefully, they are both right.  The graph (and all the graphs in the story) compare Blacks (20% of the populaton) to non-Blacks (80%).  The headline compares Black and white (60%), and the relative rate there is a bit above seven.  To find the numbers for this comparison, you just need to read the sections highlighted in red here

That’s three separate paragraphs, with a map in the middle.

Both the graph and the headline are very suited to social media (‘clickbait’ is a bit strong, since it’s actually good data). They will be circulated by lots of progressive US people who, like the lawyer whose tweet I first noticed, will look numerically clueless.   That can’t be good for the Times.

The Surgisphere papers on COVID (the chloroquine one I mentioned yesterday, and one on the safety of blood pressure drugs that affect the angiotensin system) have been retracted, at the request of three of the four authors. They had organised an independent review of the data, but:

Our independent peer reviewers informed us that Surgisphere would not transfer the full dataset, client contracts, and the full ISO audit report to their servers for analysis as such transfer would violate client agreements and confidentiality requirements. As such, our reviewers were not able to conduct an independent and private peer review and therefore notified us of their withdrawal from the peer-review process

With the combination of widespread concern about the data veracity and the unequivocal evidence from the Guardian that the Australian data were misrepresented or falsified, they didn’t have much choice.   The charitable (and quite plausible) assumption is that those three authors were dupes: someone offered them the ability to do (or at least publish) valuable analyses of COVID treatment options, and they didn’t check up on the provenance.

Ben Goldacre says this incident shows the importance of publishing code. I’m not convinced.  If I were faking a 90,000 person data set and sophisticated data analyses, I’d actually do the work of faking the data, not just the results.  Then I could publish the code perfectly safely, claim the data were confidential, but still be tripped up by someone willing to call up all the major hospitals in Australia and check whether they were involved.

Two and a half important things today.

First, a reasonably-sized, well-conducted trial of chloroquine after exposure to prevent disease found no real evidence of benefit (NYTimes) The  original research paper is supposed to be here, but isn’t up yet.

Second, Melissa Davey, Stephanie Kirchgaessner, and Sarah Boseley, in the Guardian have followed up on the concerns about that large multinational observational study.   If you missed it so far, the paper claimed to have data on over 90,000 hospitalised patients in six continents, and found patients getting chloroquine had substantially mortality. The data were supposed to have been collected from the hospitals electronically, more or less in real time, and sent to a server in the USA, but data use agreements supposedly meant that the hospitals (or even the countries they were in) couldn’t be identified — only the continents.  The Guardian journalists had the same idea as I had, that the Australian data (5 hospitals, 63 patients) was the weak point for checking: Australia is a small continent, with a small COVID outbreak, and a highly concentrated population, so you’d only have to call a fairly small number of hospitals and get them to deny sending their data to Surgisphere.  Unlike me, they actually did it:

The Guardian has since contacted five hospitals in Melbourne and two in Sydney, whose cooperation would have been essential for the Australian patient numbers in the database to be reached. All denied any role in such a database, and said they had never heard of Surgisphere. Desai did not respond to requests to comment on their statements.

So, the Australian data, after being obviously impossible and being corrected, are impossible again.  Australian patients make up only a tiny fraction of the claimed database — less than one in one thousand — but as Mies van der Rohe said, “God is in the details”

The additional half topic is that Surgisphere says they are having the data audited 

We believe that an independent academic audit that validates those three functions as it relates to our papers in the New England Journal of Medicine and The Lancet will bring further transparency to our work, further highlight the quality of our work, and also continue to deserve the confidence of our work by our colleagues.

This process will follow strict boundaries as it relates to our data use agreements, among other considerations. We are pursuing such an independent audit with all due haste while ensuring compliance with various legal and regulatory concerns.

That’s good, but they don’t say by whom. Also “This process will follow strict boundaries as it relates to our data use agreements, among other considerations” is potentially a problem. Surgisphere have claimed that the data use agreements do not allow the hospitals to be identified. An audit that does not verify the participation of at least an auditor-determined sample of the hospitals is not worth much — and ideally there would be some checks of data against records in those hospitals. A proper audit would be a lot of work, but Surgisphere claim to have carried out some such checks themselves: they say in the research paper

Collection of a 100% sample from each health-care entity is validated against financial records and external databases to minimise selection bias.

Auditing, and data access in general, are expensive and annoying if you have to do them, but so is getting the wrong answer on treatment effectiveness in COVID-19.

Team Ratings for Round 4

The basic method is described on my Department home page.
Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Performance So Far

So far there have been 24 matches played, 14 of which were correctly predicted, a success rate of 58.3%.
Here are the predictions for last week’s games.

Predictions for Round 4

Here are the predictions for Round 4. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Not for the first time in history, but it hasn’t been happening nearly enough this year.

Here’s the state of Auckland’s water storage: the width of the bars is proportional to the capacity of the dam

I got the data from aucklandwatersupply.co.nz, which uses data from Watercare to produce graphics over time.  Auckland’s total stored water today is basically the same as it was on Friday.

[Auckland also has real-time water from the Waikato and from the Onehunga spring, but having to rely on those would be bad.]