4 ms·

I looked at your prior posts and was unconvinced. From figure 1, 17% occur on highway or divided highways. However, most of the remaining categories are also t

I looked at your prior posts and was unconvinced. From figure 1, 17% occur on highway or divided highways. However, most of the remaining categories are also types of roads.

001sky 13y ago
Happy for peer review. The important point is for people to read the data and use that estimate what they perceive as the risk. At a minimum, the 33% look to be easily excluded (parking 18, residential property 6, mercantile/office property 5, open land 4). Unlcasified street is also 8%, which if it were housing association or forest roads, would be an exclusion bringing you up to 41%. Alternatively, one could just simply tighten the comps to "55mph highway", and use the 17% number as the most relevant subset. However one wants to do it, the increase in fires form (1) to (3) and the reduction of the comparison space by 1/3 to 4/5 has a pretty material change in the implied probability of a fire/event. These would be the changes vs how they appeared in October after the first PR: Observed Events (1)/Index Population%(100) October 1/100=1x Baseline November 3/100=3x Nov Min corrected data 3/67=4.5x Nov Med corrected data 3/59=5.1x Nov comps only 3/17=17.6x
- mattheww 13y ago
  Why do you want to limit by location when the the data is already separated by cause? All of the Tesla fires were caused by collision. If you compare (Fires caused by collision)/Vehicles on the road, then I agree Tesla does not compare favorably. In principle, you should compare Fires/collision in case Tesla drivers have an abnormally high collision rate. And statistically speaking, it's pretty hard to draw a conclusion when your sample size is 3 fires. But the fact is there are still no fires attributed to electrical or mechanical problems. And the points on both sides about injuries/deaths from fires are pretty much red herrings since the numbers are already extremely low. There are only 2 deaths per 100 fires caused by collisions.
- hotpockets 13y ago
  I'm also not convinced that selecting the data based on where Tesla's three fires occurred is the right why to analyze the statistics. Nor is it clear that the fires occuring in parking lots and similar areas involved solely stationary and unoccupied vehicles. In fact, all categories caused deaths at a rate similar to or greater than the fire rate except for one category, parking lots (which still caused 6% of deaths). Perhaps these fires involve someone driving too fast through a parking lot?
  - 001sky 13y ago
    The fact that Tesla was not willing to trust the Insurance companies to price insurace going forward on the same basis should be a clue that there are people revisiting their priors. And I doubt they are reading PR pieces, they may even have less ambiguous data than what we looked at here. In any event its clear tha the range from the 'rosy' projection to the 'strict' comparison is an order of magnitude in variation. It seems implausible to argue that there is 'no material change' in risk from these new data. Regardless if the issue is just a design one (and nothing really inherent with EV vs ICE).
    - hotpockets 13y ago
      Your ideas are interesting. Unfortunately you've exhausted my interest in the topic. But, I still think you overstate the absolute certainty of your interpretation. I do agree that Elon is presenting the rosier interpretation of the data.

I looked at your prior posts and was unconvinced. From figure 1, 17% occur on highway or divided highways. However, most of the remaining categories are also types of roads.

001sky 13y ago

Happy for peer review. The important point is for people to read the data and use that estimate what they perceive as the risk. At a minimum, the 33% look to be easily excluded (parking 18, residential property 6, mercantile/office property 5, open land 4). Unlcasified street is also 8%, which if it were housing association or forest roads, would be an exclusion bringing you up to 41%. Alternatively, one could just simply tighten the comps to "55mph highway", and use the 17% number as the most relevant subset. However one wants to do it, the increase in fires form (1) to (3) and the reduction of the comparison space by 1/3 to 4/5 has a pretty material change in the implied probability of a fire/event. These would be the changes vs how they appeared in October after the first PR: Observed Events (1)/Index Population%(100) October 1/100=1x Baseline November 3/100=3x Nov Min corrected data 3/67=4.5x Nov Med corrected data 3/59=5.1x Nov comps only 3/17=17.6x

mattheww 13y ago
Why do you want to limit by location when the the data is already separated by cause? All of the Tesla fires were caused by collision. If you compare (Fires caused by collision)/Vehicles on the road, then I agree Tesla does not compare favorably. In principle, you should compare Fires/collision in case Tesla drivers have an abnormally high collision rate. And statistically speaking, it's pretty hard to draw a conclusion when your sample size is 3 fires. But the fact is there are still no fires attributed to electrical or mechanical problems. And the points on both sides about injuries/deaths from fires are pretty much red herrings since the numbers are already extremely low. There are only 2 deaths per 100 fires caused by collisions.
hotpockets 13y ago
I'm also not convinced that selecting the data based on where Tesla's three fires occurred is the right why to analyze the statistics. Nor is it clear that the fires occuring in parking lots and similar areas involved solely stationary and unoccupied vehicles. In fact, all categories caused deaths at a rate similar to or greater than the fire rate except for one category, parking lots (which still caused 6% of deaths). Perhaps these fires involve someone driving too fast through a parking lot?
- 001sky 13y ago
  The fact that Tesla was not willing to trust the Insurance companies to price insurace going forward on the same basis should be a clue that there are people revisiting their priors. And I doubt they are reading PR pieces, they may even have less ambiguous data than what we looked at here. In any event its clear tha the range from the 'rosy' projection to the 'strict' comparison is an order of magnitude in variation. It seems implausible to argue that there is 'no material change' in risk from these new data. Regardless if the issue is just a design one (and nothing really inherent with EV vs ICE).
  - hotpockets 13y ago
    Your ideas are interesting. Unfortunately you've exhausted my interest in the topic. But, I still think you overstate the absolute certainty of your interpretation. I do agree that Elon is presenting the rosier interpretation of the data.