There seem to be a few threads on this already, and we've received some communications from Jason (the founder of GreenHybrid) on the subject, so I think it's time we had a stickied thread about it.

Anyone can sign up for an account and add information to the mileage database, true or otherwise. It appears that there is some erroneous information in there - impossibly low MPGs, etc. - and we're thinking of ways to lessen the impact of this erroneous data.

Here are a few things we've been thinking about:

1. Manual moderation of the database - this is something that has been done from time to time, in the past. An administrator goes in, looks for any data that appears to be incorrect (for example, a car getting 0.5MPG), and removing it by hand.
2. Discarding (or simply not processing) data that is impossibly high or low, such as that car that supposedly gets 0.5MPG.
3. Modifying the reports to include more information, such as the interquartile mean (the range of 25%-75%, so all the "outliers" are discarded.)

Manually editing the database is an OK solution, but it can be time consuming and will become moreso as more people start using the database. If you have comments on alternate solutions (such as #2 and #3 above), please feel free to post here.

Additionally, if anything in the database "jumps out at you" as being impossible, please feel free to post a link to it here, and we'll take a look at it. If everyone reports these problems in the same thread, it makes it a bit easier to track down.

Some things that can be automated by database macros:

1. Toss out anyone with only 1 or 2 tanks after a certain date duration. We know that the first tank is often meaningless because of the dealer activities, and if people are not serious about maintaining the data they should be removed.

2. Toss out anyone with totally unreasonable accumulated FE, high or low, as bogus after the same time period. The trick here is to make sure that the achievable FE at each end is maintained. Weird one-time events do happen, (such as actually taking a HiHy off-road).

3. In the interest of retaining an accurate indicator of FE to a non-hybrid owner, the cumulative FE for the class should be established without using the extreme cars in the top and bottom ranges. (i.e. Leave some percentage of the top and bottom performing cars in the published list, but eliminate them from the processing list before summing and averaging for the class cumulative FE.

Finally, there will be some cases where obvious tampering by mean-minded individuals will require hand deletion. But standards should be set so as to be fair to all concerned. Perhaps a review by two or three serious volunteers for each car class could detemine if a car gets manually deleted or not. I would take a turn at it if asked.

It is the ignorant among us that will eventually destroy us all.

Addressing your points, Stuart:

1. In the past, I had always moderated the database, but after data was already published. There was never a significant issue with fake data because I never gave it time to accumulate.

2. This would be an excellent auto-moderation idea, however that should be put on the back burner until the current data are dealt with.

3. Most people here seem to overlook that the database's overall figures are provided as medians, not means. This means that most data is already "thrown out" immediately; only the very central car of each model is displayed on the graph (which, of course, is misaligned presently).

Visit my newly-acquired website, Volconvo Debate Forums!

The first thing I'm going to do is put some limits on what people can submit. There will be limits of min/max for gallons per tank, MPG, and distance. I have the code for this about halfway done. Some maintenance is being done on the machine I'm using to develop the code, so I won't be able to finish it today, but I can move it to the live site next week.

median and IQR solve most of this already. using a trimmed mean and SD based on IQR would take care of those issues.

Please fix the dual Altima database problem by combining them or eliminating the wrong data. Please have the Altima icon at the top of the page connect to the correct fixed database. And please place altima.gif in the car icon section so we finally see our icon and not the little blue question mark. You can take it from my avatar if you don't have it.

The correct database is:
http://www.greenhybrid.com/compare/m...an-altima.html

But when you click on the Altima icon at the top you get:
http://stg.greenhybrid.com/compare/m...an-altima.html

And altima.gif needs to go here.
http://greenhybrid.com/images/cars/ It has been missing for months.

Thanks!

Done, done, and done!

There was only one guy who had good data and wasn't a staff member submitting test data, and he'd already copied his info over to the correct location.

Still working on the part that rejects crazy data.

I fixed the Prius II issue (it was showing something like 25MPG.) I'm done with the new version of the "tank adder" and ready to move it into place, after Jason reviews the limits I've established.

Cars2007.

Thanks for all you have done so far.

I'm very sorry for being a pain, but the car icons are still not consistent in what they do.

For instance:
As long as I am in "www.greenhybrid.com", the Altima icons seem to work, but every now and then, clicking on Discussion or clicking on Discussion and then the Altima icon puts me back in "stg.greenhybrid.com" and the chart is wrong and the icons don't work right.

It seems to me that your stg mirror isn't updated.

Good catch... Fixed.