Project

General

Profile

Range sizes

for analyses that require only range sizes, not full-fledged range maps, there are alternative, much faster techniques that can be used to estimate range size:

bounding-circle

  1. calculate centroid of occurrence points for the species
  2. determine each point's radius: the distance from it to the centroid
  3. remove outliers using box plot of radii1
  4. calculate the standard deviation of the radii
  5. calculate the bounding-circle radius: use 1x the standard deviation, because we want this to also be a good estimator for long, narrow ranges
    • this includes 70% of the points in the area estimate
  6. calculate the bounding-circle area, which is an estimate for the convex-hull/maxent range size

1 needed because standard deviation is a mean, not a median, and therefore sensitive to outliers
.

linear regression

  1. fit a line to the points to determine what axis, if any, they may be distributed along
  2. rotate the points so the axis is completely vertical or completely horizontal
  3. remove x- and y-outliers using box plots
  4. calculate width and height of the bounding box (x_max - x_min, etc.)
  5. calculate the area of an ellipse which fills the bounding box, which is an estimate for the convex-hull/maxent range size
  • note that this doesn't work as well for v-shaped distributions, but a convex hull won't, either
  • for circular distributions, this includes more points than the bounding-circle method because it uses the x/y-range instead of the standard deviation