Quality Percentage Question

[NinjaPablo]

How is the 'Percentage Quality Locations' calculated? I know it says "Percentage of observations with GPS coordinates and not filtered out", but what would cause an observation to be filtered out?

[uhtu]

we mark points as filtered if they exhibit anomalous behavior, being in the middle of the ocean, significantly statistically deviant from other observations of that network, etc.

[izzy4505]

How non-strange does the data have to be? I've picked up ap's as far as 8 miles away. Also, keep in mind boats can have aps.

[izzy4505]

Here's another crazy thought. . . Say somebody moves, and takes their AP with them. Somebody triangulates the new position, but it's about 80 miles off of the old data. Does the new data just get ignored?

[uhtu]

why, we had that very problem come up. :-)

we're looking into good temporal solutions (more recently seen is better, etc). finding general weighting heuristics that don't suck takes time. it will always affect the triangulation, and will just re-adjust as it is observed in the new location more frequently.

remember, WiGLE.net data is just what's been reported as being seen, not what is.
..that's a much deeper philosophical question ;->

[mycroft]

Also, when the initial "points in the ocean" problem came up, it was because of data which was obviously wrong, as in either lat or long being some default value or 0, etc. It caused lines of points extending out into the atlantic. I believe, and uhtu can smack me for talking out of turn if I'm wrong, this is the kind of "points in the ocean" data which is being filtered. At least I don't think they're calculating the lat/long of the coast at all points and saying that anything on the wrong side of the line is bad data... If they are, bobzilla is a far greater pimp than I realized.

[bobzilla]

I just put in some dead zones in the deep ocean, that cleared things up nicely.