Stats Chat

Q: Did you see  Every bite of ultraprocessed food will increase your chance of an early death.

A: How could they possibly have shown this?

Q: New study says

A: But think about it. How do they measure people’s consumption of ultraprocessed food down to the single bite level? How do they find a comparison group with just one bite less consumption? What does it even mean?

Q: Don’t I get to ask those questions?

A: <sigh>

Q: Where is the new research? And how many people or mice did they study?

A: It’s in the American Journal of Preventive Medicine, and it doesn’t involve any new data collection.  It’s a combination of a bunch of existing studies that looked at correlations between diet and health

Q: Does the claim about “every bite” appear in the journal paper?

A: No.

Q: Where does it come from?

A: As text, I’m not sure.  As an idea I think it comes from a claim in the paper for a “linear dose-response relationship”, ie, that risk goes up in a straight line with consumption of ultra-processed foods

Q: Does it?

A: That’s the model they assumed. They don’t really seem to have assessed the evidence for non-linearity in the paper, and they certainly can’t say anything meaningful about the “single bite” dose. The combined estimate is that groups of people whose consumption of these foods was higher by about 10% of their diet had a risk of premature death that was about 3% higher.

Q: But what would the risk from “one bite” be?

A:  You mean if the relationship was exactly linear extrapolated way down to that level and perfectly causal?

Q: Yes, just for fun

A: One bite of what? Different bites have different calorie levels

Q: Just work with me here. Make something up, ok?

A: Ok. So if one bite was 2% of one day’s energy intake it would be  0.0055% of one year’s energy intake, or 0.00055% of ten years’. At a increase of 3% per 10 percentage points, that’s an increased risk of premature death by a factor of 1.000016.  Which would be quite hard to detect.

Q: We were promised comparisons

A: Ah, yes. The BBC also wrote about this study, with the perfectly reasonable headline Ultra-processed foods may be linked to early death

Q: “May be“?

A: As one expert they quoted said “It’s still far from clear whether consumption of just any UPF at all is bad for health, or what aspect of UPFs might be involved. This all means that it’s impossible for any one study to be sure whether differences in mortality between people who consume different UPF amounts are actually caused by differences in their UPF consumption. You still can’t be sure from any study of this kind exactly what’s causing what.”

Q: He’s not saying ultra-processed food is good, though?

A: No, I think the most negative opinion you’ll see is just that it’s not necessarily a helpful way to categorise food

Team Ratings for Round 16

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 75 matches played, 51 of which were correctly predicted, a success rate of 68%.
Here are the predictions for last week’s games.

Predictions for Round 16

Here are the predictions for Round 16. 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 Week 17

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 128 matches played, 88 of which were correctly predicted, a success rate of 68.8%.
Here are the predictions for last week’s games.

Predictions for Week 17

Here are the predictions for Week 17. 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 23

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 154 matches played, 116 of which were correctly predicted, a success rate of 75.3%.
Here are the predictions for last week’s games.

Predictions for Round 23

Here are the predictions for Round 23. 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 Week 12

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 54 matches played, 38 of which were correctly predicted, a success rate of 70.4%.
Here are the predictions for last week’s games.

Predictions for Week 12

Here are the predictions for Week 12. 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 9

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 64 matches played, 38 of which were correctly predicted, a success rate of 59.4%.
Here are the predictions for last week’s games.

Predictions for Round 9

Here are the predictions for Round 9. 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 Week 9

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 62 matches played, 38 of which were correctly predicted, a success rate of 61.3%.
Here are the predictions for last week’s games.

Predictions for Week 9

Here are the predictions for Week 9. 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.

Stuff has a piece by Lloyd Burr on whether more New Zealanders have been having problems at the US border.  There’s a denominator problem: we want to know not just how many people are detained, but how many travellers that’s out of.

Both numbers are a bit unsatisfactory, and it’s mostly the fault of the Ministry of Foreign Affairs and Trade. We read

Last week, the Ministry of Foreign Affairs and Trade (MFAT) said eight New Zealanders had been detained at the US borders since Donald Trump came to power in November 2024, with another person arrested for “immigration-related reasons”.

A government department that pays close attention to foreign affairs might notice that Donald Trump did not in fact “come to power” in November 2024. The election was in November, but the US has an unusual policy of delaying the actual transfer of power to late January, the week before Auckland Anniversary Day.  The allegations of aggressive border enforcement didn’t start immediately after inauguration, but you might still reasonably start the clock on January 20th.

MFAT said the rate was similar to pre-Covid, which is somewhat helpful. Stuff asked (good to see!) about the comparison to last year, and got the reply that the 2023/24 financial year had nine incidents, the same that we have now seen in the shorter time since November or perhaps January.  The right comparison would depend on how many NZ citizens (since they’re the people who will get consular assistance  from MFAT) had actually travelled to the US in that time period. I don’t know that number, but MFAT should know it or at least be able to get a good estimate if they were feeling helpful.

The NZ$ Herald (from the Telegraph) has a headline Five surprising everyday medications that can help prevent dementia. This is surprising, because the general consensus is that we don’t really have any medications to prevent dementia.

The first example is the shingles vaccine. Here there’s actually some evidence — a ‘natural experiment‘ in Wales, where the roll-out of the vaccine meant people born just before and just after a particular date had different eligibility.  This isn’t quite a randomised trial, since there are other things that vary by birth date, but it’s better than the usual correlations. In particular, it’s better because the claimed effect is pretty small: a 20% reduction in risk, which is much too small to see reliably in the usual correlational studies.  This one is also good because any dementia prevention is a pure bonus; you should already want to get the shingles vaccine to prevent shingles.

Next in the story are statins, the widely-used cholesterol-lowering drugs.  Here the headline evidence is correlation, saying a reduction of 13% in dementia risk was seen in statin users in a Korean study. People have been suggesting statins as dementia prevention for at least twenty-five years and the main change is that the estimates of benefit have gotten a lot smaller.  Again, it may not matter a lot whether this is true — if you’d benefit from statins for dementia prevention you’d probably benefit from them for heart attack prevention.

Viagra is always popular in settings like this — it’s already a repurposed drug, and it’s relatively safe — but for Viagra we don’t even have correlations, just the claim that it opens up smaller blood vessels and that maybe nitric oxide (which Viagra increases) is related to memory in some way we haven’t discovered.

The trendy new drug semaglutide is on the list, again for correlations.  It could be that they really prevent dementia (and if so, we will find out from clinical trials now underway) but it’s also quite feasible that it’s just correlation.  The basis of the claim is that people who got a new expensive injected treatment in short supply for diabetes had fewer new cases of dementia in the short term than people who got a standard drug treatment. This could easily be reversing cause and effect — people with mild pre-dementia symptoms could easily be less likely to get the fancy new drugs.  In New Zealand, it doesn’t really matter if Ozempic prevents dementia, since you can’t actually get it.

And finally, the old TB vaccination that we don’t use any more is on the list. That’s actually somewhat supported by randomised trials, but the trials are in bladder cancer patients who get the BCG bacteria injected into the bladder rather than being trials of ordinary vaccination.  More research is needed before this one could really be described as “can help prevent dementia”. It’s also worth noting that we stopped routine BCG vaccination for a reason: being vaccinated means you can’t straightforwardly be tested for TB exposure.

Overall, this is another example of the “aliens” problem with dementia research.  There aren’t any breakthroughs and there are very few promising treatments. It’s a very hard problem, like detecting alien life, so the news we get is either  genuine early research — “planet hundreds of light years away is the right temperature for liquid water” — or it’s crop circles and flying saucers.

Team Ratings for Week 16

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 120 matches played, 85 of which were correctly predicted, a success rate of 70.8%.
Here are the predictions for last week’s games.

Predictions for Week 16

Here are the predictions for Week 16. 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.