Question about the hybrid database methodology
#1
Question about the hybrid database methodology
I'm curious about the methodology for computing the average fuel economy for each model, on the main graph on the front page of the "compare" tab. Are all the tanks from a particular model averaged together, or are the lifetime MPGs of each individual vehicle averaged together? I believe these two methodologies would produce significantly different results.
If you average the lifetime average MPG of all the vehicles in the database together than every vehicle will have the same influence on the average, despite the number of tanks entered, the number of miles on the car, etc.
If you average all the tanks entered into the database, the individual vehicles with more tanks added into the database will have a greater influence than the individual vehicles with fewer tanks entered into the database.
The cars seem to get the worst mileage when they are brand new. The dealers often don't fill the first tank all the way up, leading to a low first tank. They seem to break in over the first several thousand miles and get better mileage. They seem to get better mileage after the first oil change. They get better mileage as the drivers learn how to drive them.
Many of the cars in the database have very few tanks. People probably log in here when they first purchase, enter a tank or two, then lose interest when the "newness" wears off.
If you are using the "average of averages" methodology I think this may lead to an unrealistically low number.
Whoever has access to the raw numbers, can you run it both ways to see what the difference is? I'm kind of a nerd I guess, but I'd really be interested to dig into the data a little deeper and see what I can find.
If you average the lifetime average MPG of all the vehicles in the database together than every vehicle will have the same influence on the average, despite the number of tanks entered, the number of miles on the car, etc.
If you average all the tanks entered into the database, the individual vehicles with more tanks added into the database will have a greater influence than the individual vehicles with fewer tanks entered into the database.
The cars seem to get the worst mileage when they are brand new. The dealers often don't fill the first tank all the way up, leading to a low first tank. They seem to break in over the first several thousand miles and get better mileage. They seem to get better mileage after the first oil change. They get better mileage as the drivers learn how to drive them.
Many of the cars in the database have very few tanks. People probably log in here when they first purchase, enter a tank or two, then lose interest when the "newness" wears off.
If you are using the "average of averages" methodology I think this may lead to an unrealistically low number.
Whoever has access to the raw numbers, can you run it both ways to see what the difference is? I'm kind of a nerd I guess, but I'd really be interested to dig into the data a little deeper and see what I can find.
#2
Re: Question about the hybrid database methodology
Just dug into the methodology for figuring lifetime fuel economy for each car. It looks like the database is set up to average together the fuel economy of each tank to determine the lifetime fuel economy of the vehicle. This leads to a bit of error too. This gives equal influence over the lifetime average to each tank, despite the number of miles driven on the tank or the amount of fuel used.
For example: I often take long road trips. I always top off the tank right before I leave, because I don't want to stop on my trip, and gas is usually cheaper at home than at my destinations. If I only drive 100 miles, then top off, the 100 miles could easily be significantly higher or lower than a normal tank. By averaging the MPG of each tank together to get the vehicles lifetime fuel economy, you give equal weight to that short tank.
If you divide the total number of miles entered by the total number of gallons entered, you'll get a much more accurate calculation of lifetime MPG.
My car only has three tanks, and each was driven to near empty before filling, so my numbers are close using both methods. Using the "average of averages" method you have it listed at 49.8 MPG. Dividing the number of miles driven by the number of gallons consumed calculates to 49.9 MPG.
If I wanted to see my vehicle climb to the top of the rankings, I could easily do it, while being 100% accurate with my tanks. The place I like to fill up has a gradual hill, about 4 miles long, leading down to the freeway, and another gas station. I can get over 90 miles per gallon between the station I fill up, and the freeway. If I stopped at that station and topped off the tank from the 4 miles I just drove after every time I actually fill up, every other tank would be 90 MPG. even though this tank was only four miles, it carries the same weight as all my 600 mile "full" tanks.
A lot of people fill up when they find a lower price on gas, regardless of how much gas they have left. The lifetime average could be significantly off for them, or for anyone who fills up with varying amounts of gas left before empty.
Anyone follow that?
For example: I often take long road trips. I always top off the tank right before I leave, because I don't want to stop on my trip, and gas is usually cheaper at home than at my destinations. If I only drive 100 miles, then top off, the 100 miles could easily be significantly higher or lower than a normal tank. By averaging the MPG of each tank together to get the vehicles lifetime fuel economy, you give equal weight to that short tank.
If you divide the total number of miles entered by the total number of gallons entered, you'll get a much more accurate calculation of lifetime MPG.
My car only has three tanks, and each was driven to near empty before filling, so my numbers are close using both methods. Using the "average of averages" method you have it listed at 49.8 MPG. Dividing the number of miles driven by the number of gallons consumed calculates to 49.9 MPG.
If I wanted to see my vehicle climb to the top of the rankings, I could easily do it, while being 100% accurate with my tanks. The place I like to fill up has a gradual hill, about 4 miles long, leading down to the freeway, and another gas station. I can get over 90 miles per gallon between the station I fill up, and the freeway. If I stopped at that station and topped off the tank from the 4 miles I just drove after every time I actually fill up, every other tank would be 90 MPG. even though this tank was only four miles, it carries the same weight as all my 600 mile "full" tanks.
A lot of people fill up when they find a lower price on gas, regardless of how much gas they have left. The lifetime average could be significantly off for them, or for anyone who fills up with varying amounts of gas left before empty.
Anyone follow that?
Last edited by RockMonkey; 04-29-2007 at 10:19 PM.
#3
Re: Question about the hybrid database methodology
On the vehicle scale, I believe the individual records are broken into four equal sized groups. Then the vehicle averages use the center pair. Thus the best and worst vehicle records are dropped for the fleet average.
Within a single vehicle, it is possible to use either the individual calculated MPGs or the miles and gallons. So far, I've not found a significant difference with either method. In the case of the Prius, we know the tanks have a variable capacity. What this means is some tanks have excellent MPG and other are poor. However, the MPG averages over time seem to even out.
One last comment, MPG is not a competitive sport except within the performance of each vehicle. Improving my vehicle performance pays me a dividend. Being better or worse than another vehicle has no payback at all.
Bob Wilson
Within a single vehicle, it is possible to use either the individual calculated MPGs or the miles and gallons. So far, I've not found a significant difference with either method. In the case of the Prius, we know the tanks have a variable capacity. What this means is some tanks have excellent MPG and other are poor. However, the MPG averages over time seem to even out.
One last comment, MPG is not a competitive sport except within the performance of each vehicle. Improving my vehicle performance pays me a dividend. Being better or worse than another vehicle has no payback at all.
Bob Wilson
#4
Re: Question about the hybrid database methodology
Here's some clarification:
Each individual vehicle is given a lifetime fuel economy figure calculated by total miles / total gallons. The model figure is then calculated using the median of the vehicles in that category. That way, a single car (no matter how drastic the outlier) will not skew the figure, a car with 100,000 miles will not matter more than a car with 50,000 and the final figure is as accurate as possible.
I put a lot of thought into it, and the median is the best way to go.
Each individual vehicle is given a lifetime fuel economy figure calculated by total miles / total gallons. The model figure is then calculated using the median of the vehicles in that category. That way, a single car (no matter how drastic the outlier) will not skew the figure, a car with 100,000 miles will not matter more than a car with 50,000 and the final figure is as accurate as possible.
I put a lot of thought into it, and the median is the best way to go.
Thread
Topic Starter
Forum
Replies
Last Post