Interesting study. I’m no statiscian but the number of ebike crashes vs escooter crashes anaylzed is wildly different - 35 vs 686. The article says it’s still statistically significant, but I can’t help feel weird about comparing 3 ebike crashes and 156 escooter crashes in Liverpool.
Judging by their stats they must have had data on 150,000-200,000 e bike trips overall.
So if you treat each trip as an relevant event for determining the crash rate - then I think it’s a decent enough sample size of eBike trips. It’s just they had heck of a lot more of eScooter data.
The main biassed I’d think are around the rental companies they got data from, and the customer populations.
And that they have basically only 7 cities.
but the findings were remarkably consistent over 3 measures of exposure and within each city except Dusseldorf (not enough ebikers to have any crashes).
A couple of the p-values were over 5% or 10% , so some were weaker when narrowing down to an city or just one of the exposure measures - but still a fairly consistent pattern.
Interesting study. I’m no statiscian but the number of ebike crashes vs escooter crashes anaylzed is wildly different - 35 vs 686. The article says it’s still statistically significant, but I can’t help feel weird about comparing 3 ebike crashes and 156 escooter crashes in Liverpool.
Judging by their stats they must have had data on 150,000-200,000 e bike trips overall. So if you treat each trip as an relevant event for determining the crash rate - then I think it’s a decent enough sample size of eBike trips. It’s just they had heck of a lot more of eScooter data.
The main biassed I’d think are around the rental companies they got data from, and the customer populations. And that they have basically only 7 cities.
but the findings were remarkably consistent over 3 measures of exposure and within each city except Dusseldorf (not enough ebikers to have any crashes).
A couple of the p-values were over 5% or 10% , so some were weaker when narrowing down to an city or just one of the exposure measures - but still a fairly consistent pattern.