croupier 1.2.0 → 1.3.0

data/README.md

@@ -32,7 +32,16 @@ where:
 
 ### Available distributions and options
 
-Current version provides: Cauchy, exponential, normal, Poisson, triangular and uniform distributions.
+Current version implements the following distributions: 
+
+* Cauchy
+* Exponential
+* Geometric
+* Negative binomial
+* Normal
+* Poisson
+* Triangular
+* Uniform
 
 To get a list of all the available distributions use the ```--help``` (or ```-h```) option with croupier:
 

data/lib/croupier/distributions/geometric.rb

@@ -0,0 +1,45 @@
+module Croupier
+  module Distributions
+
+    #####################################################################
+    # Geometric Distribution
+    # Discrete probability distribution that expresses the number of X
+    # Bernoulli trials needed to get one success, supported on the
+    # set { 1, 2, 3, ...}
+    #
+    # Wikipedia -- http://en.wikipedia.org/wiki/Geometric_distribution
+    # I made this choice because it's Knuth choice.
+    # (The Art of Computer Programming, Volume 2, 3.4.1.F )
+    class Geometric < ::Croupier::Distribution
+
+      def initialize(options={})
+        @name = "Geometric distribution"
+        @description = "Discrete probability distribution that expresses the the number of X Bernoulli trials needed to get one success, supported on the set { 1, 2, 3, ...}"
+        configure(options)
+        raise Croupier::InputParamsError, "Probability of success must be in the interval [0,1]" if params[:success] > 1 || params[:success] < 0
+      end
+
+      # Fair point: it is not the inverse of the cdf,
+      # but it generates the distribution from an uniform.
+      def inv_cdf n
+        (Math.log(1-n) / Math.log(1-params[:success])).ceil
+      end
+
+      def default_parameters
+        {:success => 0.5}
+      end
+
+      def self.cli_name
+        "geometric"
+      end
+
+      def self.cli_options
+        {:options => [
+           [:success, 'success probability of each trial', {:type=>:float, :short => "-p", :default => 0.5}]
+         ],
+         :banner => "Geometric distribution. Discrete probability distribution that expresses the number of X Bernoulli trials needed to get one success, supported on the set { 1, 2, 3, ...} }"
+        }
+      end
+    end
+  end
+end

data/lib/croupier/distributions/nbinomial.rb

@@ -0,0 +1,57 @@
+module Croupier
+  module Distributions
+
+    #####################################################################
+    # Negative Binomial Distribution
+    # Discrete probability distribution of the number of successes in a
+    # sequence of Bernoulli trials before a specified (non-random)
+    # number of failures (denoted size) occur.
+    #
+    # The parameter prob expresses the probability of success.
+    # Wikipedia -- http://en.wikipedia.org/wiki/Negative_binomial_distribution
+    class Nbinomial < ::Croupier::Distribution
+
+      def initialize(options={})
+        @name = "Negative binomial distribution"
+        @description = "Discrete probability distribution of the number of successes in a sequence of Bernoulli trials before a specified (non-random) number of failures (denoted size) occur."
+        configure(options)
+        raise Croupier::InputParamsError, "Probability of success must be in the interval [0,1]" if params[:success] > 1 || params[:success] < 0
+      end
+
+      # Fair point: it is not the inverse of the cdf,
+      # but it generates the distribution from an uniform.
+      def generate_sample n=1
+        generate_geometrics(n).each_slice(params[:size]).map do |sample|
+          sample.inject(-params[:size], &:+) # Inject starts on -size because
+          # this way it is equivalent to:
+          # sample.map{|x| x - 1}.inject(&:+)
+        end
+      end
+
+      def base_geometric
+        ::Croupier::Distributions::Geometric.new(success: params[:success])
+      end
+
+      def generate_geometrics(n)
+        base_geometric.generate_sample(params[:size]*n)
+      end
+
+      def default_parameters
+        {:success => 0.5, :size => 1}
+      end
+
+      def self.cli_name
+        "nbinomial"
+      end
+
+      def self.cli_options
+        {:options => [
+           [:size, 'number of errors', {:type => :integer, :default => 1}],
+           [:success, 'success probability of each trial', {:type=>:float, :short => "-p", :default => 0.5}]
+         ],
+         :banner => "Negative binomial distribution. Discrete probability distribution of the number of successes in a sequence of Bernoulli trials before a specified (non-random) number of failures (denoted size) occur."
+        }
+      end
+    end
+  end
+end

data/test/distributions/R_tests/test_geometric.R

@@ -0,0 +1,23 @@
+context('** Geometric Distribution **')
+
+# It uses the fact that it is equivalent to a negative binomial
+# distribution with parameters size=1 and prob = p (Wikipedia
+# set p_{nbinomial} to be the fail probability while R takes
+# it as the success one).
+# Besides, since for the equality to hold the support must start
+# at zero, it just substracts one from the vector of samples.
+context('ChiSquare Goodness of Fit test for p=0.5')
+  croupier_geometric <- read.table("../generated_samples/geometric_05.data")
+  cs_result<-summary(goodfit(croupier_geometric$V1 - 1, type="nbinomial", method="ML", par=list(size=1, prob=0.5)))
+
+  test_that("p-value > 0.05, geometric, p = 0.5", {
+    expect_true(cs_result['Pearson',3] > 0.05)
+  })
+
+context('ChiSquare Goodness of Fit test for p = 0.05')
+  croupier_geometric <- read.table("../generated_samples/geometric_005.data")
+  cs_result<-summary(goodfit(croupier_geometric$V1 - 1, type="nbinomial", method="ML", par=list(size=1, prob=0.05)))
+
+  test_that("p-value > 0.05, geometric, p=0.05", {
+    expect_true(cs_result['Pearson',3] > 0.05)
+  })

data/test/distributions/R_tests/test_nbinomial.R

@@ -0,0 +1,17 @@
+context('** Negative Binomial Distribution **')
+
+context('ChiSquare Goodness of Fit test for prob=0.5, size=5')
+  croupier_nbinomial <- read.table("../generated_samples/nbinomial_05_5.data")
+  cs_result<-summary(goodfit(croupier_nbinomial$V1, type="nbinomial", method="ML", par=list(size=5, prob=0.5)))
+
+  test_that("p-value > 0.05, nbinomial, prob = 0.5, size=5", {
+    expect_true(cs_result['Pearson',3] > 0.05)
+  })
+
+context('ChiSquare Goodness of Fit test for prob = 0.25, size=15')
+  croupier_nbinomial <- read.table("../generated_samples/nbinomial_025_15.data")
+  cs_result<-summary(goodfit(croupier_nbinomial$V1, type="nbinomial", method="ML", par=list(size=15, prob=0.25)))
+
+  test_that("p-value > 0.05, nbinomial, prob=0.25, size=15", {
+    expect_true(cs_result['Pearson',3] > 0.05)
+  })

data/test/distributions/R_tests/test_normal.R

@@ -7,11 +7,11 @@ context('Kolmogorov-Smirnov test for default mean = 0 and std = 1')
   croupier_normal <- read.table("../generated_samples/normal_0_1.data")
   ks_result<-ks.test(croupier_normal$V1, "pnorm")
 
-  test_that("p-value > 0.05", {
+  test_that("p-value > 0.05, normal (0,1)", {
     expect_true(ks_result$p.value > 0.05)
   })
 
-  test_that("statistic converging to 0", {
+  test_that("statistic converging to 0, normal(0,1)", {
     expect_true(as.numeric(ks_result$statistic) < 0.05)
   })
 
@@ -19,11 +19,11 @@ context('Shapiro test for default mean = 0 and std = 1')
   croupier_normal <- read.table("../generated_samples/normal_0_1.data")
   s_result<-shapiro.test(croupier_normal$V1[1:5000])
 
-  test_that("p-value > 0.05", {
+  test_that("p-value > 0.05, normal (0,1)", {
     expect_true(s_result$p.value > 0.05)
   })
 
-  test_that("statistic converging to 0", {
+  test_that("statistic converging to 0, normal(0,1)", {
     expect_true(as.numeric(ks_result$statistic) < 0.05)
   })
 
@@ -31,11 +31,11 @@ context('Kolmogorov-Smirnov test for given mean = 5 and std = 6')
   croupier_normal <- read.table("../generated_samples/normal_5_6.data")
   ks_result<-ks.test((croupier_normal$V1 - 5)/6, "pnorm")
 
-  test_that("p-value > 0.05", {
+  test_that("p-value > 0.05, normal(5,6)", {
     expect_true(ks_result$p.value > 0.05)
   })
 
-  test_that("statistic converging to 0", {
+  test_that("statistic converging to 0, normal(5,6)", {
     expect_true(as.numeric(ks_result$statistic) < 0.05)
   })
 
@@ -43,10 +43,10 @@ context('Shapiro test for default mean = 5 and std = 1')
   croupier_normal <- read.table("../generated_samples/normal_0_1.data")
   s_result<-shapiro.test((croupier_normal$V1[1:5000] - 5 ) / 6)
 
-  test_that("p-value > 0.05", {
+  test_that("p-value > 0.05, normal(5,6)", {
     expect_true(s_result$p.value > 0.05)
   })
 
-  test_that("statistic converging to 0", {
+  test_that("statistic converging to 0, normal(5,6)", {
     expect_true(as.numeric(ks_result$statistic) < 0.05)
   })

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: croupier
 version: !ruby/object:Gem::Version
-  version: 1.2.0
+  version: 1.3.0
   prerelease: 
 platform: ruby
 authors:
@@ -10,7 +10,7 @@ authors:
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2012-06-21 00:00:00.000000000 Z
+date: 2012-06-22 00:00:00.000000000 Z
 dependencies: []
 description: Croupier is a Ruby gem to generate a random sample of numbers with a
   given probability distribution.
@@ -30,6 +30,8 @@ files:
 - lib/croupier/distribution.rb
 - lib/croupier/distributions/cauchy.rb
 - lib/croupier/distributions/exponential.rb
+- lib/croupier/distributions/geometric.rb
+- lib/croupier/distributions/nbinomial.rb
 - lib/croupier/distributions/normal.rb
 - lib/croupier/distributions/poisson.rb
 - lib/croupier/distributions/triangular.rb
@@ -38,6 +40,8 @@ files:
 - lib/croupier.rb
 - test/distributions/R_tests/test_cauchy.R
 - test/distributions/R_tests/test_exponential.R
+- test/distributions/R_tests/test_geometric.R
+- test/distributions/R_tests/test_nbinomial.R
 - test/distributions/R_tests/test_normal.R
 - test/distributions/R_tests/test_poisson.R
 - test/distributions/R_tests/test_triangular.R
@@ -76,6 +80,8 @@ summary: Samples of random numbers with specific probability distributions
 test_files:
 - test/distributions/R_tests/test_cauchy.R
 - test/distributions/R_tests/test_exponential.R
+- test/distributions/R_tests/test_geometric.R
+- test/distributions/R_tests/test_nbinomial.R
 - test/distributions/R_tests/test_normal.R
 - test/distributions/R_tests/test_poisson.R
 - test/distributions/R_tests/test_triangular.R