Average MPG graphs misleading

[jostlehim]

The average mpg bar graphs for a car model are misleading, due to the discontinuous nature of the classes. As an example, if you go to http://www.fuelly.com/car/ford/fiesta you will see what appears to be a straightforward distribution of mpgs. If you look to the right hand side of the graph, you see that the range is discontinuous, going 47,48,51,52,58.

This is because bars are only drawn if you have 2 or more values in that category. However, this is very misleading. What is worse, if you filter on Diesel L4, you can see that there are also entries for 56,61,62,64 and 66 with a further 60 under Diesel H4 and a 67 under Diesel (no specification).

So, the graph totally obscures the fact that there is skewed distribution. The problems this can result in is very clear when you select the 2010 Fiestas. This looks like a continuous distribution, but there is a huge gap in the middle, and it is actually two separate distributions, due to the difference between Petrol and Diesel fuel economies. This graph is simply misleading.

By all means adjust the classes to deal with low numbers (ie graph 41-45, 46-50, 51-55) but the graphs should have equal class sizes, and should include all values.

[DTMAce]

I can say from looking at them periodically, they were designed to round the averages to the nearest whole MPG, and only show those that have been reported. As for having 2 or more, that was to eliminate mistaken MPG importing or possible fakes. I mean if say most people have a MPG of between 20-30 for a car and one person has 50 just once, then that person may have made a mistake in their data entry. So to minimize this, only ranges with 2 or more entries will be reported.

The point of them are to show you where most people fall, not give you a full representation of every possible MPG range in general.

I look at the caravan's list, I see many are right where mine is. That tells me that my van is getting the MPG that most of the people reporting are. That is all I need to know.

Fuelly is not about being an exacting site, but an averaging site. It has good reporting information but may not be to everyone's standards.

I am simply one of many members, I like how it is reported. I use an excel sheet locally for more exact and other relevant info that I wish to know that fuelly does not support. Best part is, I simply import the data right from Fuelly so I only have to input it once.

Hope this makes sense.

[mathowie]

Yeah, sorry about the confusion on the graphs, but like DTMAce says, it's really just approximating averages and our graphing toolset is really limited so we're not going for absolute statistical accuracy because we can't really do it given the constraints of what's available for graphing engines.

[BDC]

If they include all the values, then we have to look at that one guy who doesn't understand how to use Odometer tracking and adds in 150mpg tanks every now and then.

If the Fuelly.com admins wanted to run a super-accurate data collection and distribution site, they would need to spend a heck of a lot of time making both the collection and the distribution systems way more accurate. Obviously it is not meant to be a data collection site, which is why Fuelly is super easy to use to record your own mileage, but takes a little time to figure out how to find the data that you want. Fortunately it appears that you are smart enough to find the data that you want, so no problem, right? And for a free site with no advertising, we're actually lucky that they've given us as much access to site data as they have.

[jostlehim]

My issue is not so much that classes where n

I appreciate that you are not looking for absolute statistical accuracy, but this is not even approximate accuracy. The way to deal with outliers would be to restrict the graph to ±2 standard deviations from the mean.

How about putting a filter on fuel type (Diesel, all configurations|Petrol, all configurations) which gets around the confusion people seem to have getting a consistent engine configuration. Incidentally, the FAQ currently says that H4 refers to hybrid engines - I thought it stood for horizontal as opposed to inline? Nearly all of the Subarus are down as H4 or H6, and they aren't hybrid, but Boxers....

[pb]

Yeah, Subaru calls their engines "Horizontal" and so the H is doing double-duty for that particular context. Only Subarus have that, but I understand the confusion. We might need to add a separate engine type just for them if/when they start using hybrid engines.

[pb]

I added a note to the FAQ about the Subaru H4 and H6 engine types.

[jostlehim]

Porsche also have flat or boxer engines and indeed the Porsche Boxster, 911, Carrera and Cayman are down as having an H6

http://en.wikipedia.org/wiki/Flat_engine

[pb]

yeah, we might need to clarify some of those down the road. I think it's ok having H mean different things in different contexts for now. We'll just have to be sure to explain it carefully. Thanks for pointing that out.

[vr4Legacy]

Subaru also offers Turbo in every class but the Tribeca... so a T4 or H4T would be a GREAT addition.... if/when it get's added. Or maybe add an option... # cylinders, then engine orientation (i.e. Hybrid, Turbo, Naturally Aspirated, Diesel, Other)