Module: Histogram

Included in:
Array, NArray
Defined in:
lib/histogram/plot.rb,
lib/histogram/version.rb,
lib/histogram.rb

Defined Under Namespace

Modules: Plot

Constant Summary collapse

VERSION = 
"0.2.2.0"
DEFAULT_BIN_METHOD = 
:scott
DEFAULT_QUARTILE_METHOD = 
:moore_mccabe

Class Method Summary collapse

Instance Method Summary collapse

Class Method Details

.iqrange(obj, opts = {}) ⇒ Object

opts:

defaults:
:method => :moore_mccabe, :tukey
:sorted => false



52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
 
# File 'lib/histogram.rb', line 52

def iqrange(obj, opts={})
  opt = {:method => DEFAULT_QUARTILE_METHOD, :sorted => false}.merge( opts )
  srted = opt[:sorted] ? obj : obj.sort
  sz = srted.size
  return 0 if sz == 1
  answer = 
    case opt[:method]
    when :tukey
      hi_idx = sz / 2
      lo_idx = (sz % 2 == 0) ? hi_idx-1 : hi_idx
      median(srted[hi_idx..-1]) - median(srted[0..lo_idx])
    when :moore_mccabe
      hi_idx = sz / 2
      lo_idx = hi_idx - 1
      hi_idx += 1 unless sz.even?
      median(srted[hi_idx..-1]) - median(srted[0..lo_idx])
    else
      raise ArgumentError, "method must be :tukey"
    end
  answer.to_f
end

.median(sorted) ⇒ Object

finds median on a pre-sorted array



75
76
77
78
 
# File 'lib/histogram.rb', line 75

def median(sorted)
  return sorted[0] if sorted.size == 1
  (sorted[(sorted.size - 1) / 2] + sorted[sorted.size / 2]) / 2.0
end

.minmax(obj) ⇒ Object

returns (min, max)



17
18
19
20
21
22
23
24
25
26
27
28
29
 
# File 'lib/histogram.rb', line 17

def minmax(obj)
  if obj.is_a?(Array)
    obj.minmax
  else
    mn = obj[0]
    mx = obj[0]
    obj.each do |val|
      if val < mn then mn = val end
      if val > mx then mx = val end
    end
    [mn, mx]
  end
end

.sample_stats(obj) ⇒ Object

returns (mean, standard_dev) if size == 0 returns [nil, nil]



33
34
35
36
37
38
39
40
41
42
43
44
 
# File 'lib/histogram.rb', line 33

def sample_stats(obj)
  _len = obj.size
  return [nil, nil] if _len == 0
  _sum = 0.0 ; _sum_sq = 0.0
  obj.each do |val|
    _sum += val
    _sum_sq += val * val
  end
  std_dev = _sum_sq - ((_sum * _sum)/_len)
  std_dev /= ( _len > 1 ? _len-1 : 1 )
  [_sum.to_f/_len, Math.sqrt(std_dev)]
end

Instance Method Details

#avg_ints(one, two) ⇒ Object

:nodoc:



350
351
352
 
# File 'lib/histogram.rb', line 350

def avg_ints(one, two) # :nodoc:
  (one.to_f + two.to_f) / 2.0
end

#histogram(*args) ⇒ Object

Returns [bins, freqs]

histogram(bins, opts) histogram(opts)

Options:

:bins => :scott    Scott's method    range/(3.5σ * n^(-1/3))
         :fd       Freedman-Diaconis range/(2*iqrange *n^(-1/3)) (default)
         :sturges  Sturges' method   log_2(n) + 1 (overly smooth for n > 200)
         :middle   the median between :fd, :scott, and :sturges
         <Integer> give the number of bins
         <Array>   specify the bins themselves

:bin_boundary  => :avg      boundary is the avg between bins (default)
                  :min      bins specify the minima for binning

:bin_width => <float> width of a bin (overrides :bins)
:min => <float> # explicitly set the min
:max => <float> # explicitly set the max val

:other_sets => an array of other sets to histogram

Examples

require 'histogram/array'
ar = [-2,1,2,3,3,3,4,5,6,6]
# these return: [bins, freqencies]
ar.histogram(20)                  # use 20 bins
ar.histogram([-3,-1,4,5,6], :bin_boundary => :avg) # custom bins

# returns [bins, freq1, freq2 ...]
(bins, *freqs) = ar.histogram(30, :bin_boundary => :avg, :other_sets => [3,3,4,4,5], [-1,0,0,3,3,6])
(ar_freqs, other1, other2) = freqs

# histogramming with weights
w_weights.histogram(20, :weights => [3,3,8,8,9,9,3,3,3,3]) 

# with NArray
require 'histogram/narray'
NArray.float(20).random!(3).histogram(20)
   # => [bins, freqs]  # are both NArray.float objects

Notes

  • The lowest bin will be min, highest bin the max unless array given.

  • Assumes that bins are increasing.

  • :avg means that the boundary between the specified bins is at the avg between the bins (rounds up )

  • :min means that to fit in the bin it must be >= the bin and < the next (so, values lower than first bin are not included, but all values higher, than last bin are included. Current implementation of custom bins is slow.

  • if other_sets are supplied, the same bins will be used for all the sets. It is useful if you just want a certain number of bins and for the sets to share the exact same bins. In this case returns [bins, freqs(caller), freqs1, freqs2 …]

  • Can also deal with weights. :weights should provide parallel arrays to the caller and any :other_sets provided.



169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
 
# File 'lib/histogram.rb', line 169

def histogram(*args)
  make_freqs_proc = lambda do |obj, len|
    if obj.is_a?(Array)
      Array.new(len, 0.0)
    elsif obj.is_a?(NArray)
      NArray.float(len)
    end
  end

  case args.size
  when 2
    (bins, opts) = args
  when 1
    arg = args.shift
    if arg.is_a?(Hash)
      opts = arg
    else
      bins = arg
      opts = {}
    end
  when 0
    opts = {}
    bins = nil
  else
    raise ArgumentError, "accepts no more than 2 args"
  end

  opts = ({ :bin_boundary => :avg, :other_sets => [] }).merge(opts)

  bins = opts[:bins] if opts[:bins]
  bins = DEFAULT_BIN_METHOD unless bins

  bin_boundary = opts[:bin_boundary]
  other_sets = opts[:other_sets]

  bins_array_like = bins.kind_of?(Array) || bins.kind_of?(NArray) || opts[:bin_width]
  all = [self] + other_sets 

  if bins.is_a?(Symbol)
    bins = number_of_bins(bins)
  end

  weights = 
    if opts[:weights]
      have_frac_freqs = true
      opts[:weights][0].is_a?(Numeric) ? [ opts[:weights] ] : opts[:weights]
    else
      []
    end

  # we need to know the limits of the bins if we need to define our own bins
  if opts[:bin_width] || !bins_array_like 
    calc_min, calc_max = 
      unless opts[:min] && opts[:max]
        (mins, maxs) = all.map {|ar| Histogram.minmax(ar) }.transpose
        [mins.min, maxs.max]
      end
  end
  _min = opts[:min] || calc_min
  _max = opts[:max] || calc_max

  if opts[:bin_width]
    bins = []
    _min.step(_max, opts[:bin_width]) {|v| bins << v }
  end

  _bins = nil
  _freqs = nil
  if bins_array_like
    ########################################################
    # ARRAY BINS:
    ########################################################
    _bins = 
      if bins.is_a?(Array)
        bins.map {|v| v.to_f } 
      elsif bins.is_a?(NArray)
        bins.to_f
      end
    case bin_boundary
    when :avg
      freqs_ar = all.zip(weights).map do |xvals, yvals|

        _freqs = make_freqs_proc.call(xvals, bins.size)

        break_points = [] 
        (0...(bins.size)).each do |i|
          bin = bins[i]
          break if i == (bins.size - 1)
          break_points << avg_ints(bin,bins[i+1]) 
        end
        (0...(xvals.size)).each do |i|
          val = xvals[i]
          height = have_frac_freqs ? yvals[i] : 1
          if val < break_points.first
            _freqs[0] += height
          elsif val >= break_points.last
            _freqs[-1] += height 
          else
            (0...(break_points.size-1)).each do |i| 
              if val >= break_points[i] && val < break_points[i+1]
                _freqs[i+1] += height
                break
              end
            end
          end
        end
        _freqs 
      end
    when :min
      freqs_ar = all.zip(weights).map do |xvals, yvals|

        #_freqs = VecI.new(bins.size, 0)
        _freqs = make_freqs_proc.call(xvals, bins.size)
        (0...(xvals.size)).each do |i|
          val = xvals[i]
          height = have_frac_freqs ? yvals[i] : 1
          last_i = 0
          last_found_j = false
          (0...(_bins.size)).each do |j|
            if val >= _bins[j]
              last_found_j = j
            elsif last_found_j
              break
            end
          end
          if last_found_j ; _freqs[last_found_j] += height ; end
        end
        _freqs
      end
    end
  else
    ########################################################
    # NUMBER OF BINS:
    ########################################################
    # Create the scaling factor

    dmin = _min.to_f
    conv = _max == _min ? 0 : bins.to_f/(_max - _min)

    _bins = 
      if self.is_a?(Array)
        Array.new(bins)
      elsif self.is_a?(NArray)
        NArray.float(bins)
      end

    freqs_ar = all.zip(weights).map do |xvals, yvals|

      # initialize arrays
      _freqs = make_freqs_proc.call(xvals, bins)
      _len = size

      # Create the histogram:
      (0...(xvals.size)).each do |i|
        val = xvals[i]
        height = have_frac_freqs ? yvals[i] : 1
        index = ((val-_min)*conv).floor
        if index == bins
          index -= 1
        end
        _freqs[index] += height
      end
      _freqs
    end

    # Create the bins:
    iconv = 1.0/conv
    case bin_boundary
    when :avg
      (0...bins).each do |i|
        _bins[i] = ((i+0.5) * iconv) + dmin
      end
    when :min
      (0...bins).each do |i|
        _bins[i] = (i * iconv) + dmin
      end
    end
  end
  [_bins] + freqs_ar
end

#number_of_bins(methd = DEFAULT_BIN_METHOD, quartile_method = DEFAULT_QUARTILE_METHOD) ⇒ Object

returns(integer) takes :scott|:sturges|:fd|:middle

middle is the median between the other three values

inspired by Richard Cotton’s matlab implementation and the histogram page on wikipedia



89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
 
# File 'lib/histogram.rb', line 89

def number_of_bins(methd=DEFAULT_BIN_METHOD, quartile_method=DEFAULT_QUARTILE_METHOD)
  if methd == :middle
    [:scott, :sturges, :fd].map {|v| number_of_bins(v) }.sort[1]
  else
    nbins = 
      case methd
      when :scott
        range = (self.max - self.min).to_f
        (mean, stddev) = Histogram.sample_stats(self) 
        range / ( 3.5*stddev*(self.size**(-1.0/3)) )
      when :sturges
        1 + Math::log2(self.size)
      when :fd
        2 * Histogram.iqrange(self, :method => quartile_method) * (self.size**(-1.0/3))
      end
    nbins = 1 if nbins.nan?
    nbins = 1 if nbins <= 0
    nbins.ceil.to_i
  end
end