[elif]

If a seatbelt makes you 50% safer in the event of a potentially fatal crash, you wouldn't say "it's really a 4% difference because 8% of the population is involved in a potentially fatal accident"

That is purposefully obfuscating the relevant data.

[brokenkebab]

The wording you used for your example is purposely unwieldy, but even so it would be still logically valid. It's perfectly valid as well to say that it obfuscates a particular point you are trying to make. However, as humans tend to have different priorities, and values a wording you prefer may be validly presented as obfuscating points other people see as more important.

[elif]

It is a perfectly reasonable example. Both are individual decisions one makes for individual safety considerations. Introducing society-wide multipliers makes it harder for an individual to make decisions based upon the data. When I get in a car, I am not buckling up for society. When I invest food, I am not making a decision for society.

Unless you believe the primary function of this research is to inform policy decisions which have little to no effect, and not consumers, who have a very direct effect?

[brokenkebab]

I'm sure your intentions are good, but what I wrote is that any logically/mathematically truthful statement is, well, true. The fact that it emphasizes this, or that side of the argument doesn't make it false, and its value is in the eye of beholder, because proper desicions are based on cost/benefits assessment where both costs, and benefits differ between people/societies/tribes/whatever.

And what you offer me to discuss in response is that your position in your particular example is certainly right.

[DanBC]

Risk does need to be presented both ways - relative and absilute - for anyone to make any sense of it though.

People see 50% increase in risk and don't have enough information to make an informed choice. If you tell them that we go from 2 people in a thousand suffering ill effects to 3 people in a thousand they can make their choice.

See the work of Gerd Gigerenzer for plenty of examples of people caused harm because they were told only the relative risk.

[elif]

But in this case, the extra variable adds weird implications that are simply not true...

For instance, if half of the population became vegetarians, the rate of colorectal cancer would go down, and red meat would become even more "safe."

[DanBC]

People aren't saying that.

They're saying if you have a population of 1,000 people who don't eat red meat you'd expect to see 2 cases of DISEASE_X. If you take a similar population but who do eat red meat you'd expect to see 3 cases of DISEASE_X.

That's a 50% increase, but that 50% relative risk number isn't enough information for people to make any decision. Here we're talking about an extra 1 person in 1000 people. But if the numbers were 4 in 10 increasing to 6 in 10 that's still only a 50% increase.

We need the absolute risk numbers to get an understanding: 0.2% increases to 0.3%. Something that's unlikely to happen becomes marginally more likely to happen.

This is not some weird obscure trickery. It's an important thing for people receiving healthcare to understand. It's mainstream science.

https://bestpractice.bmj.com/info/toolkit/practise-ebm/under...

https://www.eufic.org/en/understanding-science/article/absol...

Have a look at the factbox and the icon array linked in this page: https://www.harding-center.mpg.de/en/fact-boxes/early-detect...

But also, we should avoid using percentages and we should use natural frequencies.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3310025/ Only about 1 in 4 people understand that 0.1% means 1 in 1000.

[parineum]

It depends on the audience.

Personally, if something will make a low occurence event marginally safer, I might not care. That's what this paper is saying.

Perhaps a scientific paper isn't the place for that (an article about the paper probably is) but that translated data is still relevant to a certain perspective.

[Sawamara]

But this is exactly the point of studies like this. They are written in a way that allows misinterpretation, and then the media can pick it up and run with it (like they did with this one).

There is a large body of writing about P-hacking, bias in studies, this is just one more in that long, long list.