What makes our database different is that we try to indicate the reliability of sources and the confidence we have in the co-ordinates which are published on the site. This allows our readers to make a better assessment of the information available, and in the long term allows them to put greater trust in the database.
SITE CO-ORDINATE CONFIDENCE RATING SYSTEM
All sites in the OpenAtlas have a 'confidence rating' which allows readers to assess if the co-ordinate and therefore map position of the site in question is correct. The ranking goes from 0 (confidence unknown) to 10 (confirmed by GPS). In general rankings from 8 and above can be relied upon to be precise and accurate, whereas anything below 5 is likely to be incorrect.
GAZETTEER RELIABILITY SYSTEM
Our database uses a system to judge the probable 'reliability' of any particular source and therefore its co-ordinates, and makes this available in our 'Search for Sites' tool. Reliability is calculated from two terms: 'accuracy' and 'precision'. A higher reliability rating means that the co-ordinate has a greater chance of being closer to the real location of the site.
Accuracy is a judgement given on the sources' normal accuracy as a percentage, thus 0% would be never accurate and 100% would be always accurate: a database created from GPS collection should be 99% accurate for example, since they are verified, or taken from direct GPS co-ordinates, though there is always the possiblity of input error. Most databases probably fall between the 80-95% range.
Precision is a known or typical average precision of the source in metres (thus ASPRO whose co-ordinates are never more than 5 km precise, will have a precision value of '5000' metres), a GPS-based database will have a precision of c 2-8 m and so a precision of '4' m would be appropriate.
Reliability is thus a factor of these both and can be calculated by the following formula:
maximprecision = 40000
reliability = ((maximprecision - precision)/maximprecision) * accuracy
The 'maximum imprecision' is again a matter of judgement, since for different purposes different level of precisions are appropriate. However, for most ArchAtlas needs, anything which is less precise than 40km is considered to be too imprecise. Negative values of 'reliability' are thus possible but indicate that the source should not be used.
The resultant value of reliability is 'percentage-like' but should not be considered to reflect a percentage of any particular measurement.