FileMaker Go: Location, Part 2: What’s Accuracy?

This is Part 2 in a multi-part series about using the location functions in FileMaker Go.

  1. Where Am I?
  2. What’s Accuracy?
  3. Where’s My Stuff?
  4. How Far?
  5. Which Way?
  6. What’s Close?

In my last post, I described the basics of using the Location and LocationValues functions in FileMaker Go. Now I want to focus on something that’s been bugging me for a while about these functions: accuracy.

GPS is pretty good at pinpointing your location, but there are several likely sources of error to account for, so we can’t realistically expect it to be exactly right — thus the accuracy component of a location. Accuracy is accepted by both the Location and the LocationValues functions as a parameter, and accuracy is returned by both functions as a result. But what exactly does FileMaker mean by “accuracy”?

There are a few different measures of GPS accuracy in common use, including CEP, R95, and 2DRMS. CEP and R95 accuracies specify that your true position is within the given radius of the returned coordinates a certain percentage of the time, 50% and 95%, respectively. 2DRMS is twice the expected root mean squared error of the returned coordinates from your true position. None of FileMaker’s documentation describes what measure of accuracy the location functions are using. Neither does Apple’s documentation of the CLLocation class.

In the absence of documentation, I set out to figure it out for myself. I built an accuracy testing file to repeatedly get location samples from the same spot to see what kind of variability we get for different accuracy values. I asked other folks to help gather more data in more locations.

Accuracy Testing

I don’t have a whole lot to go on yet, but here’s what I’ve figured out so far:

FileMaker is not reporting CEP, R95, 2DRMS, or anything else consistent. For some reported accuracies, FileMaker appears to be reporting the 90% confidence radius (true position is within that radius 90% of the time); for others, it looks like the 75% confidence radius. For some positions, the reported accuracy is orders of magnitude bigger than the actual variation in the coordinate positions.

If FileMaker’s “accuracy” is smaller, the coordinates are probably closer to your true position. There just isn’t an obvious pattern of how much closer. Series of location values with 5 or 10 meter reported accuracies are closer together than locations with 30 or 65 meter accuracies. Getting any more detailed than that gets messy fast.

It isn’t practically useful to ask for better than 10 meter accuracy. With smaller accuracy values, the reported location is usually closer to your true position, but only up to a point. The variations in locations with 5 and 10 meter accuracies are about the same. If you’re comparing a location with an accuracy of 5 meters with another with a 10 meter accuracy, you may as well flip a coin to decide which is closer to the true position.

FileMaker’s accuracy is more about confidence than variability. If locations with 5 and 10 meter accuracies have about the same variation, why are those location fixes being reported with different accuracies? Another anomaly is that locations on my WiFi-only (no GPS) iPad always report a 65 meter accuracy (about the diameter of a typical WiFi base station’s coverage area), but the variation in reported coordinates is an order of magnitude smaller than that. My guess is that iOS and FileMaker report a smaller accuracy number when they are more confident in the fix they have. (In statistics nerd terms, the accuracy distance does not reliably act like it’s a threshold for a frequency probability, leaving me to presume that it should be treated as a belief probability threshold.)

Accuracy is a blunt measure. I’ve seen the nice, round numbers of 5, 10, 30, 50, and 65 meter accuracies, but never anything in-between like 42 or 6.789. If FileMaker and iOS were trying to make the returned accuracy value exactly whatever it is it’s supposed to be, I’d expect to see more nuance in the numbers. iOS is probably taking a computational shortcut for performance purposes that’s good enough for most consumer applications. Smaller is better, but don’t take the exact number too seriously.

I didn’t get as much sample data from other folks in the FileMaker community as I hoped. If you could download the accuracy testing file to your iOS devices and record some sample data for me, I’d really appreciate it. Maybe more data will lead me to better — possibly even completely different — answers. AccuracyTesting.fmp12

2016-11-01T17:01:19+00:00 October 22nd, 2013|