Optimal Battery Size and Driving Habits

[FANGO]

I know that's what you would think, but the label on Figure 2 is incorrect by a factor of 10x. It should say .05, meaning 5%.

Furthermore, Figure 2 isn't cumulative, so 5% (in that statistic) is just the average percentage of vehicles-days exactly between 50 miles and 60 miles, not the percentage with 50 miles to more.

And in addition to that, very few people would buy a car based on average single day miles. They want to know: how many times per week, per month, per quarter, or per year, would they have to rent a car (or do something else). For example, if the percentage of days with more than 60 miles is just 17% (in that statistic), this could still mean that it might be more than 60% who have such a trip once or more often per month. That is something that report (and the underlying data) doesn't tell. But that would be the more important number, I'd say.

You can stop it with the patronizing, it's pretty obvious that those are percentages, and what a histogram is.

A histogram isn't an "average." Nowhere are "averages" ever mentioned in my portion of this discussion. Of course I already said this, and yet you still persist in talking about averages. Which is why I've been uninterested in this for a long time.

Luckily this is its own thread now, so I can unsubscribe instead of listen to people pretending to teach me how to read a histogram.

 

[Norbert]

You can stop it with the patronizing, it's pretty obvious that those are percentages, and what a histogram is.

A histogram isn't an "average." Nowhere are "averages" ever mentioned in my portion of this discussion. Of course I already said this, and yet you still persist in talking about averages. Which is why I've been uninterested in this for a long time.

Luckily this is its own thread now, so I can unsubscribe instead of listen to people pretending to teach me how to read a histogram.

When it says 17% of the days are with 60 miles or more, than that is an average per vehicle. It does't tell you whether all those days come from the same 17% of the vehicles (and the others have none of those days), or if all of the vehicles have a few of those days. It doesn't tell you what the distribution is. It is a single statistical number, which doesn't tell you enough about the individual cases.

Are you able to confirm that its 5% and not 0.5%, in Figure 2 of the report you quote?

 

[Norbert]

For example, if each and every vehicle made a more-than-60-miles trip every 6 days, that would result in a number of about 17% as well. But it would mean that 100% would want a car with more than 60 mile range, if they can have such a car for a similar price.

 

[Norbert]

So I downloaded the NHTS data and did some more analysis on it. A lot more than what you can query with the online query tool.
What I got was 3.4% of each household had a vehicle with 100+ miles per day.

The way the data was collected was the subjects had 1 week of journal entries of their estimated miles/time (some were exact and some were rounded) and there was a few problematic data points (either that or they can drive 195 mph for an entire day).
I think this may skew the data also as I would assume there would be more traveling in certain times of the year than others (non-school days).

This data really shows that people have lots of short trips during the day.

So for a primary vehicle 100 miles/day is tough for 3.4% of the households.

Are you sure the data is meant to have 1 week for each subject? ... guess I have to write code to check for that.

Update: I'm deleting my previous message with the question about the report's SQL since it turns out that the database has only (and exactly) 1 day for each vehicle.

I guess it wasn't that nice as I found a mistake in the way I was grouping before. Households that travel >100 miles/day have multiple days >100 miles in there sample weeks.

Sorry, but there are no "sample weeks" in DAYV2PUB. The data is for one day only, for each vehicle.


This means that the percentages I have calculated are for single days, not for 'at least one day in a week'. This means the extrapolation for one week is higher than the percentages I have given. The same applies to the longer-term extrapolations.

But before posting new results, I'll make more double-checks (and try to include the weighting).

 

[Norbert]

The question now is this: If an average of about 6% of vehicles drive 100 or more miles on randomly chosen days, how many can be expected to have at least one such day in a week (or month, quarter, year) ?

 

[Norbert]

I think the weighting should be done in this way:

In general (without considering the weights), for each vehicle, a 1 or 0 is added, if the sum of trips for its test-day is larger or equal to a certain milage.

Now, there are vehicle weights and travel day weights in that database. While the report (quoted above) uses only travel day weights, I think that each 1 and 0 should be weighted by the combination of vehicle weight and travel day weight. The combined weight is the 2 weights multiplied with each other. And then used in the same way as a single weight is used normally.

As an example, without weights, if there are 2 vehicles, one with a large mileage, and one with a small mileage, the calculation would be (1 + 0) / 2 = 0.5 (50%)

With weights: Let's say the first vehicle weight is 5, the first travel day weight is 4, the second weights would be 7 and 2.
Then the combined weights would be 5*4 = 20 and 7*2 = 14.

then the calculation would be (20 * 1 + 14 * 0) / (20 + 14). Which is 20 / 34 = 0.588 (59%).

Objections? (Not that I am expecting any, still documenting my efforts.)

 

[aronth5]

The stats posted are not averages, and we aren't talking about averages, and I don't believe I have mentioned overall daily mileage averages at any point in this discussion. The stats say how often people drive over 150 miles per day. It is well less than 1%.

In fact, only .7% of cars are driven over 100 miles in any given day, as established previously in this thread, by me and by another poster. That means twice a year give or take. So 150 is even less than that. And 250 is even less...in fact it's literally off the chart.

These questions are all answered in the statistics.

Too much over analysis and reliance on stats. There are enough owners and prospective owners that want and whose driving experience would improve with greater range that Tesla will increase it. The only question is how much and what is the timeframe? Verified by Elon who stated at today's annual meeting that "you should expect improvement in range over time."