Tested on latest MRI 1.9.x, 2.x, and jruby (no NArray support for jruby). Known to work on MRI 1.8.7.
require 'histogram/array' # enables Array#histogram data = [0,1,2,2,2,2,2,3,3,3,3,3,3,3,3,3,5,5,9,9,10] # by default, uses Scott's method to calculate optimal number of bins # and the bin values are midpoints between the bin edges (bins, freqs) = data.histogram # equivalent to: data.histogram(:scott, :bin_boundary => :avg)
Multiple types of binning behavior:
# :scott, :fd, :sturges, or :middle data.histogram(:fd) # use Freedman-Diaconis method to calc num bins data.histogram(:middle) # (median value between the three methods) (bins, freqs) = data.histogram(20) # use 20 bins (bins, freqs) = data.histogram([-3,-1,4,5,6]) # custom bins (bins, freqs) = data.histogram(10, :min => 2, :max => 12) # 10 bins with set min and max # bins are midpoints, but can be set as minima (bins, freqs) = data.histogram([-3,-1,4,5,6], :bin_boundary => :min) # custom bins with :min # can also set the bin_width (which interpolates between the min and max of the set) (bins, freqs) = data.histogram(:bin_width => 0.5)
Sometimes, we want to create histograms where the bins are calculated based on all the data sets. That way, the resulting frequencies will all line up:
# returns [bins, freq1, freq2 ...] (bins, *freqs) = set1.histogram(30, :other_sets => [[3,3,4,4,5], [-1,0,0,3,3,6]])
Histograms with weights/fractions:
# histogramming with weights data.histogram(20, :weights => [3,3,8,8,9,9,3,3,3,3])
Works with NArray objects
require 'histogram/narray' # enables NArray#histogram # if the calling object is an NArray, the output is two NArrays: (bins, freqs) = .float(20).random!(3).histogram(20) # bins and freqs are both NArray.float objects
gem install histogram
Big thanks to those who have made contributions!
- deal with zero std (Greg Dean)
- support for 1.8.7 and jruby (Kiera Radman)
- reorder nan checks to fix zero stdev arrays (arjun810)