Hacker News Reader

Comment by vkou | Hacker News Reader

You've posted a case study of 'How to lie with statistics'.

This is a chart of incidents, but it doesn't define what an incident is (did anyone get hurt? how many people were hurt? how badly? was it a near-miss, or did the many layers of redundancy in safety handle it fine?), and it doesn't normalize it to incident/flight or incident/mile.

Does Malaysia Airlines having one airplane shot down by the Russians, and another sent into the drink by a pilot-mass-murder-suicide count as an incident? Are either of them Boeing's fault? Are anti-SAM missile defenses supposed to be a standard feature on commercial airliners?

The MAX crashes obviously were, and the MAX has a very bad safety record as a result.

The claim (Boeing is worse than Airbus) is probably correct, but that chart is horrible evidence for it, or for how much worse one is, compared to the other. If you're going to provide evidence for this claim, you need to find something better.

View on HN

The other problem with that chart is that its from the NTSB and not normalized to relative fleet size. Boeing aircraft are more common in the NTSB's jurisdiction.

I'm not trying to lie and I'm aware my data isn't perfect, but I'm the only one providing any data at all so far. I have repeatedly asked for better data and been denied. I can only analyze the data I have, imperfect though it may be. When someone is consistently asking for better data, criticizing that person's data is kinda dumb.

> Does Malaysia Airlines having one airplane shot down by the Russians, and another sent into the drink by a pilot-mass-murder-suicide count as an incident? Are either of them Boeing's fault? Are anti-SAM missile defenses supposed to be a standard feature on commercial airliners?

You're making the mistake of assuming Boeing's fault is what matters. If, for whatever reason, Boeing's planes are more likely to be shot down - regardless of whether or not they are at fault - they are still more dangerous. We are not at the why stage yet, we are at the what stage.

Anyway, since then I've found better data that actually implies Boeing is safer than Airbus.

> Boeing aircraft have a lower frequency of crashes per million flights compared to Airbus, with Boeing having a rate of one crash per 184 million flights and Airbus having a rate of one crash per 081 million flights.

https://www.mightytravels.com/2024/04/comparing-airline-inci...

I've also found a better breakdown comparing the events between the plane manufacturers from my initial graph.

https://statwonk.com/are-boeing-planes-more-dangerous-than-a...

> If, for whatever reason, Boeing's planes are more likely to be shot down - regardless of whether or not they are at fault - they are still more dangerous.

No reasonable person would conclude that.

You either need enough data for it to be statistically significant (Which, given the sample size of one isn't it...)

Or you can work around the lack of statistics, by having a bullet-proof explanation for why Boeings (N=1) get shot down more frequently than Aribuses (N=0). But the bar for that is much higher than 'Well, statistically speaking..."

Statistics are great, because they are a lossy shortcut inform our decisionmaking when we don't actually understand the root causes. But they require statistically significant data to do so. In the absence of it, we have to actually understand the root causes.

> You either need enough data for it to be statistically significant

I guess this wasn't clear, but I thought this was implied. When I said it's "more likely" I meant you can quantify it. You are correct of course.

Also I agree with everything else you're saying in principle. I kinda feel like we just started the conversation with a misunderstanding about shared assumptions and likely agree with each other.

Reply on news.ycombinator.com