I couldn't find any better place so i throw it in here:
Somehow I cannot believe them... (Some nutjob throwing this around claiming how EV's are better and its getting on my nerve.)
Whats the communities opinion on this?
Not going to lie - I read that and thought you were saying somebody was going around claiming electron-volts (eV) are better than volts. Took me a good hour before it suddenly dawned on me (
). It's just turned 0800 and I've been up for three hours, you'd think I'd be awake by now
I'm afraid this isn't my field - but on a first-principles basis (and looking at the presented data), there seem to be some major flaws with this somewhat preordained conclusion that the packs can be used for another few hundred thousand miles.
- Sample size - 286 samples isn't that large; may not be representative.
- Sampling bias - the samples were self-selected; may not be an issue here, but always worth remembering.
- Survivorship bias - the samples shown (see charts
[1],
[2],
[3] and
[4] in article) - how many people are dropping out, how many people are replacing their electric car when the battery dies (rather than replacing the battery), etc etc. The graph only shows the "surviving" car owners so to speak.
- Data obfuscation (probably accidental) - there are numerous occasions where the same user has submitted multiple entries: this is not intrinsically bad, as it allows you to track and individual vehicle - provided it is dealt with correctly. In this dataset, I am unsure how the data is being treated - if it is being treated as individual data points instead of an individual sampling, then this can hide poor performance [5]. If it has been dealt with properly (which I
think would entail using only the most recent entry from a given user), then I believe it's fine - if not then it may be hiding poor performance.[6]
There are a bunch of other problems that I can think of, but I won't go into detail for risk of this post becoming inordinately long(er); but suffice to say I am not convinced the observations support the article's conclusion that the batteries are on average capable of another 150'000 miles...
And it is for a much simpler line of reasoning: if they
were capable of another 150'000 miles (on average)... people would be reporting that already... unless what you're trying to say is that people get to MAX-150'000 miles (on average) and then suddenly stop and say to themselves "Hey! Let's get a new battery/car!" (though I wouldn't put it past them) - the fact that on average people
aren't getting an extra 150'000 miles out their cars suggests that there are confounding factors at play asides from the battery.
In fact, you can see this in charts [1-4] - you'll notice that there are fewer samples at the "high mileage" end of the chart... it's almost as if, on average, high-mileage electric cars don't fare as well!
[1-4] - see
google docs file for original data
[5] - As a comparison, imagine measuring the same resistor for tolerance multiple times and then claiming the entire batch of resistors is within tolerance.
[6] - As a quick demonstration: imagine two users - UserA has their car for 7 days, and then it dies. He logs the mileage the day it dies. UserB has their car for a year: he logs the mileage every day. Collating the data without filtering makes it look like UserB was in fact 365 different cars, each with n+1 days on the clock, when it was in fact just a single car. While the average is similar 186 vs 181 respectively), the conclusion is not - especially once we introduce more people - lets say UserC-UserZ: who in this example lose their cars after exactly 1 month and log exactly once. Clearly UserB was an outlier, but his excessive logging has swamped UserA and UserC-UserZ's data, driving the average higher than a single reading would have. (if UserB had made a single reading, the average would be 42-days until car-death with a standard deviation of 66: if UserB made a reading every day, the average life of a car would be ~172 days, with a standard deviation of 108, which clearly isn't the representing the life of most cars, as most cars in this sample fail at the 30 day mark; one could spend a while analysing the differences through statistical analysis - but, I'm afraid I can't be bothered... I also can't remember how to do statistical analysis properly anymore...
).
--EDIT-- some dummy couldn't count, I had to fix his calculations because they were borked