GPS Accuracy–Where’s Your GPS Been?

It occurred to me I should start my GPS calibration measurements with a simple stationary test. From the specification literature for my unit I suspected the distribution of data points around the true location might be normal. A normal distribution is a statistic term for the typical bell-shaped graph of data with tails on both sides–

So the easiest way to test this idea is to set the GPS down in one spot for awhile and let it record a location point every second to its track log. I took my Garmin 60CSx out to the backyard yesterday evening, put it down on a little table in the middle of the yard, reset it, and went for a bike ride. 3612 seconds later (1 hour and 12 seconds) I turned off the log and transferred the data to my computer. The GPS told me it had been sitting still for 57 minutes and moving for 3 minutes. This is interesting since it was stationary for the entire time. It had also decided it had traveled 381 feet. I imported the data into Topofusion and captured the trace of my wandering GPS. Click on the image to see th full-sized trace. Note that the scale line in the lower right is 5 feet–

I’ll do some quantitative analysis of the data points in Numbers (iWorks’ spreadsheet), but that doesn’t look to me like what I would expect from a normal distribution. That should look like a mound of points collected in the center with decreasing numbers further out. This looks more like a “drunken sailor” distribution. But let me do some analysis (I’ll share the data file if you want to do some…). Topofusion decided the GPS had traveled 386 feet as it sat quietly.

This reinforces the point that you need to be a little cautious with data you collect from a GPS. It’s a decent approximation of reality, but it’s not quite real. All points were within 15 feet of one another. This was in an open area with about 8 or so good satellite signals.