Introduction
This tool takes a set of static location fix data from GPSTest 4.0+, or csv files containing the same columns, and calculates the resulting point using basic averaging and Kalman filtering. It can also use log files created by Google's GnssLogger and older versions of GPSTest, but these truncate the location data to low resolution, and are not recommended.
Limitations
This tool intentionally cannot handle dynamic data; it is intended to calculate a single position only, so the device should be kept static while taking readings. It cannot create tracks from a moving device.
The data can be spread across a maximum of 4294967296 files (hopefully enough!), each up to 9007199254740991 B in size, with a maximum of 9007199254740991 data records in total. (The data record limitation affects only the mean and standard deviation calculations; Kalman filtering has no limitation.) It can draw charts for the first 4294967296 data records. Even if you somehow have enough RAM to cope with loading files of these sizes, enough storage space to actually store files of these sizes, and a device capable of producing this vast amount of data, your computer's processing power will grind to a halt long before you reach these numbers, and your browser will probably complain that script execution is taking too long. Your browser may also impose other limitations.
Standard deviation is calculated using Welford's method, as it works using a single pass, has no data size limitation, can give out instantaneous values for all currently processed data, and gives more accurate results than most other methods when the differences between readings are very small (ie. when you have a very precise GNSS device).
Note that this tool (like almost all tools) does not work well in locations very near to (at) the poles, where the longitude can have any value at all. For that, you would need an equivalent tool that works in cartesian coordinates, not geodetic.