16 New Regional Codes
There may actually be more new regional codes represented in the YTree this update than this, because I added around 200 new regional codes only found on the MTree to my geocodes database last month.
- AR-S Argentina (Santa Fe)
- AZ-AGA Azerbaijan (Ağstafa)
- AZ-SAK Azerbaijan (Şəki)
- DJ-AS Djibouti (Ali Sabieh)
- FR-47 France (Lot-et-Garonne)
- GB-HNS United Kingdom (Hounslow)
- GM-U Gambia, The (Upper River)
- KH-12 Cambodia (Phnom Penh)
- LK-11 Sri Lanka (Kŏḷamba)
- LR-MY Liberia (Maryland) - First sample from Liberia
- MA-AZI Morocco (Azilal)
- MA-ZAG Morocco (Zagora)
- OM-BU Oman (Al Buraymī)
- RS-27 Serbia (Prizrenski okrug)
- SD-8 Sudan (An Nīl al Abyaḑ)
- TR-08 Turkey (Artvin)
Minimum Radius now applied to Regions below 10,000 km2 instead of 40,000 km2
In order to improve the accuracy of the heatmaps, I'm taking into account the area of each region above 10,000 km2. Previously any region below 40,000 km2 was treated as a point with the same radius. Now a sample from a region that is 40,000 km2 will have a radius twice as big as one from a region that is 10,000 km2 (ratio of square roots).
I'm optimizing these changes for the new Y and M Heatmaps to be released next update.
M Heatmap is still under development
Y Heatmap will be updated with a more accurate calculation at the same time as M Heatmap is introduced, planned for next YFull version.
The more accurate calculation comes as a result of dividing two absolute frequency surfaces by one another.
The current (old) method I have been using approximates relative frequency by weighting samples according to sample rate of the center that, according to the heatmap.js plugin, then had to be equally distributed in each direction from the center.
The reality is that each point doesn't automatically translate into a relative frequency triangularly distributed equally in each direction away from the center - it varies depending on absolute sample rate in each direction.