Planning to make a regression model... anyone else?
 

Over the past few days I've been contemplating how much average speed affects gas mileage. So I've decided to conduct an experiment.

Turns out that my GPS has a feature that allows me to calculate moving average MPH over a given period. So from now on, every time I fill my tank I will record the miles, MPG, and the average speed for that tank. After I have enough data (probably 30 - 40 tanks), I'll make a linear regression model to see what the effect of average speed is on MPG.

(I know that MPG is not exactly linearly related to speed, but I'm just gonna force the data)

Anyone else want to do something like this? I'd be especially interested in getting a cross-section of different types of vehicles, including an SUV/pickup (F250, Suburban, that kind of thing) if there is anyone here who has one. I'll do the calculations, I just need you to do the record keeping.

(Of course the data would be messed up if you make any vehicle modifications, etc., so don't plan on making any changes. And obviously I'd need some way for you to calculate a moving average speed.)

-Bob C.

Wouldn't a ScanGauge or similar system allow you to examine speed vs mpg more directly? Or are you thinking this would be more accurate?

That sounds like a whole lot of work.

You'll have to alternate tanks at low and high average speed to avoid interference from seasonal temperature/weather changes.

I don't have a scangauge so I can't comment about what it can (or can't) do. If it can do fancy calculations like average speed per tank and whatnot, that would be cool. I'm not keen on instantaneous readings, I don't think they are accurate enough.

HC, I have a few different routes I can take to work (typical commute is 1 hour each direction) with varying speeds and distance, and when you mix in other types of driving there should be enough variability to cover a decent range of average speeds. The regression calculation itself is fairly simple (for me).

-BC

ScanGauge does indeed provide average [speed, MPG, and some others I don't remember] per tank, per day, and per trip.

I've seen one person put average speed in their gaslog but I don't remember who.

When I had a Scangauge II I used the trip cost feature (enter the price/gallon) to figure out what was the fastest way to the grocery store (two ways, about a 45 minute trip). One was a mile shorter, but the terrain varied a lot making it less than obvious which would be best. Of course I had to consider both ways, so it was four separate measurements. The shorter route won for both directions and was cheaper by about 25 cents/trip (at the nearly $4/gallon gas price at the time).

i may not be the only one, but i do log avg mph when i remember. just my dd however...

https://www.gassavers.org/garage/viewgaslog/1239

Mine works out to;

80MPG at 40 MPH
70MPG at 50 MPH
62MPG at 60 MPH
54MPG at 70 MPH

regards
gary

I log avg speed on my F250. I have an hour meter installed that turns on with the ignition 12 volt source. This way I am also averaging the time I sit at stop signs idling. It isn't all that easy to figure out what average is best. It mainly shows me how much city driving I've been doing. If I had to make a guess, about 46 mph would be my best average speed, since that is when my overdrive finally kicks in.

Why just do linear regression fitting? There are a lot of other stastical models that should be able to give you a better model. You are going to need at least "something quadratic" or possibly cubic to model any remaining third-order effects.