First the simple question: What unit does the gps accuracy use? meters? It is not mentioned in any documentation I found, e.g. here:
or here:
I want to find a reasonable threshold for accuracy like suggested in the link above with 15, but what is actually reasonable? My last survey had no bounds and had a median accuracy of 10 and mean accuracy of 75, so some high outliers, but since I do not know it what accuracy actually means I have no way of setting a reasonable boundary.
So what does it actually mean, like statistically? I could not find a solid unambigious definition online (neither on survey solutions, nor wikipedia, nor elsewhere) other than it is the circle radius around the given position representing some kind of error. Is it using something like the 99%-quantile of distances to each measured point?
I want to approximatly find an adress in a rural area with the gps coordinates and an adress will also be given, so if I can say with high certainty that the position will be in 50m of the given location that would be totally fine.
Maybe this is the wrong place to ask the second question, I do not know. Any help is appriciated
Regarding your first question on the unit of GPS accuracy: Survey Solutions records GPS accuracy in meters.
For your second question on a reasonable accuracy threshold, there is no universally fixed threshold, as this depends on several factors—including the devices used in the survey, the topography and forest cover of the study area, and the specific objectives of the survey. The survey team should determine the acceptable accuracy level based on these considerations.
You may implement a validation rule on the GPS question—either as a warning or an error. For example:
var_gps.Accuracy <=10
Alternatively, you can set accuracy thresholds in the server settings or in the Interviewer App settings.
For more details, please refer to the following documentation:
That helps me. If possible, for future users it would be helpful if the unit of measurement is mentioned in the “GPS questions” article as to not let anybody guess (it is mentioned for distances in the examples but not for the accuracy).
About the second question: I am aware that each use case has their own threshold. That is not my problem. I wanted to ask the technical definition of the gps accuracy, so how to interpret the results I get.
Circular Error Probability (CEP): CEP is a standard measure used to express GPS accuracy. It defines the radius within which 50% of all GPS measurements fall. For example, a 50% CEP of 5 meters means that 50% of the time, the GPS position will be within 5 meters of the true location. Additionally, other measures used are DRMS (Distance Root Mean Square), which defines the diameter within which 68% of the points fall, and 2DRMS, which defines the diameter within which 95% of the points fall.
So which kind of accuracy of these is actually used in Survey Solutions (50%, 68%, 95%, something else)?
If the answer is device depent I might aswell not even set an accuracy as the results are not comparable as each interviewer will get different results, which would be important to know as we are not using the same type of device for each interviewer.
There is no specific logic on application level regarding accuracy calculation.
Application runs on top of the Android operation system that supplies measurements.
Android uses underlying hardware of consumer device for retrieval.
By default, it’s 68% of confidence.
Take into consideration, that other sources as cell towers and Wi-FI could affect the final measurement (FusedLocationProvider in Android); as well as obstacles like buildings (reflected signal), device velocity (Doppler shift) could also be a factor even for the same device.
Depending on your needs, you may control accuracy in several places (device level and expressions in questionnaire).
If 50 meters is sufficient for you - use it. It could be a soft validation in case if for some reasons an interviewer is not able to get it.
Very small numbers (single digits) for accuracy could not be achievable in some circumstances (obstructed satellites, for example) and energy draining (more time for acquisition).
