Page 3
Mikhail Kanevski, Vadim Timonin
Sliding Window Analysis of Power Law Distributions.
“WinPoΛo”
Version 1.
“One-page” user manual
IGAR Research Report
No. G01-2008
The development was supported in part by Swiss National Science Foundation Projects: “GeoKernels” N 200021-113944 and “ClusterVille” N 100012-113506
Lausanne, April, 2008
What’s for ?
WinPoΛo module is designed to visualize and estimate the parameters of power law distributions.
or, in log-scale
So, the equation of the line of c.d.f. in log scale gives the parameters as
and
Usage of the module
· Start WinPoΛo module.
· Load training dataset. Number of points of the dataset is appeared in the Size field.
· Select Function ( X ) variable (if necessary).
Histogram of selected variable will appear in area 3, c.d.f. in area 4.
Both charts:
Zoom to area: with left mouse button, select area in the chart to zoom in (rectangle from top-left corner to down-right corner).
Reset zoom: with left mouse button, select any area in the chart (rectangle from top-right corner to down-left corner).
Button - copy chart (as picture) to clipboard.
Histogram chart
1. Number of bins
2. log scale (on/off)
3. Graph of Gaussian approximation of the histogram (on/off)
4. Marks (number of points in each bin label) (on/off)
Legend: number of items is equal to number of bins; values are the thresholds for ones.
c.d.f. chart
1. Graph of c.d.f. (data and Gaussian approximation) (on/off)
2. Graph to fit line (data or Gaussian approximation)
3. Scale of the graphs (lin-lin, lin-log, log-log)
Shadow area (in both charts) is a window (part of the data) was the parameters (line equation) are estimated. This area can be changed by dragging the lines on the chart. Result (fitted line equation in form of ) will appear in the title of the chart. The size (number of points) of the window is indicated as well.
Fit to data:
Fit to Gaussian approximation:
HAPPY WinPoΛoing !!!