statsample 0.6.5 → 0.6.7

data/lib/statsample/test/levene.rb

@@ -0,0 +1,90 @@
+module Statsample
+  module Test
+    # = Levene Test for Equality of Variances
+    # From NIST/SEMATECH:
+    # <blockquote>Levene's test ( Levene, 1960) is used to test if k samples have equal variances. Equal variances across samples is called homogeneity of variance. Some statistical tests, for example the analysis of variance, assume that variances are equal across groups or samples. The Levene test can be used to verify that assumption.</blockquote>
+    # Use:
+    #   require 'statsample'
+    #   a=[1,2,3,4,5,6,7,8,100,10].to_scale
+    #   b=[30,40,50,60,70,80,90,100,110,120].to_scale
+    # 
+    #   levene=Statsample::Test::Levene.new([a,b])
+    #   puts levene.summary
+    #   
+    # Output:
+    #   Levene Test
+    #   F: 0.778121319848449
+    #   p: 0.389344552595791
+    #
+    # Reference:
+    # * NIST/SEMATECH e-Handbook of Statistical Methods. Available on http://www.itl.nist.gov/div898/handbook/eda/section3/eda35a.htm
+    class Levene
+      # Degrees of freedom 1 (k-1)
+      attr_reader :d1
+      # Degrees of freedom 2 (n-k)
+      attr_reader :d2
+      # Name of test
+      attr_accessor :name
+      # Input could be an array of vectors or a dataset
+      def initialize(input, opts=Hash.new())
+        @vectors=input
+        @name="Levene Test"
+        opts.each{|k,v|
+          self.send("#{k}=",v) if self.respond_to? k
+        }
+        compute
+      end
+      # Value of the test
+      def f
+        @w
+      end
+      
+      def to_reportbuilder(generator) # :nodoc:
+        generator.add_text(summary)
+        
+      end
+      # Summary of results
+      def summary
+        "#{@name}
+F: #{f}
+p: #{probability}"
+      end
+      
+      def compute
+        n=@vectors.inject(0) {|ac,v| ac+v.n_valid}
+        
+        zi=@vectors.collect {|vector|
+          mean=vector.mean
+          vector.collect {|v| (v-mean).abs }.to_scale
+        }
+        
+        total_mean=zi.inject([]) {|ac,vector|
+          ac+vector.valid_data
+        }.to_scale.mean
+      
+        k=@vectors.size
+        
+        sum_num=zi.inject(0) {|ac,vector|
+          ac+(vector.size*(vector.mean-total_mean)**2)
+        }
+        
+        sum_den=zi.inject(0) {|ac,vector|
+          z_mean=vector.mean
+          ac+vector.valid_data.inject(0) {|acp,zij|
+            acp+(zij-z_mean)**2
+          }
+        }
+        @w=((n-k)*sum_num).quo((k-1)*sum_den)
+        @d1=k-1
+        @d2=n-k
+      end
+      private :compute
+      # Probability.
+      # With H_0 = Sum(s2)=0, probability of getting a value of the test upper or equal to the obtained on the sample
+      def probability
+        1-Distribution::F.cdf(f, @d1, @d2) 
+      end
+      
+    end
+  end
+end

data/lib/statsample/test/umannwhitney.rb

@@ -24,7 +24,8 @@ module Statsample
       # Parameters:
       # * n1: group 1 size
       # * n2: group 2 size 
-      # Reference: Dinneen, L., & Blakesley, B. (1973). Algorithm AS 62: A Generator for the Sampling Distribution of the Mann- Whitney U Statistic. Journal of the Royal Statistical Society, 22(2), 269-273
+      # Reference: 
+      # * Dinneen, L., & Blakesley, B. (1973). Algorithm AS 62: A Generator for the Sampling Distribution of the Mann- Whitney U Statistic. <em>Journal of the Royal Statistical Society, 22</em>(2), 269-273
       # 
       def self.u_sampling_distribution_as62(n1,n2)
 
@@ -98,14 +99,18 @@ module Statsample
       attr_reader :r1
       # Sample 2 Rank sum
       attr_reader :r2
-      # Sample 1 U
+      # Sample 1 U (useful for demostration)
       attr_reader :u1
-      # Sample 2 U
+      # Sample 2 U (useful for demostration)
       attr_reader :u2
-      # U Value
+      # U Value 
       attr_reader :u
-      # Compensation for ties
+      # Value of compensation for ties (useful for demostration)
       attr_reader :t
+      #
+      # Create a new U Mann-Whitney test
+      # Params: Two Statsample::Vectors
+      # 
       def initialize(v1,v2)
         @n1=v1.valid_data.size
         @n2=v2.valid_data.size
@@ -128,6 +133,7 @@ module Statsample
         @u2=r2-((@n2*(@n2+1)).quo(2))
         @u=(u1<u2) ? u1 : u2
       end
+      # Report results.
       def summary
         out=<<-HEREDOC
 Mann-Whitney U
@@ -141,8 +147,11 @@ Z: #{sprintf("%0.3f",z)} (p: #{sprintf("%0.3f",z_probability)})
         end
           out
       end
-      # Exact probability of finding values of U lower or equal to sample on U distribution. Use with caution with m*n>100000
-      # Reference: Dinneen & Blakesley (1973)
+      def to_reportbuilder(generator) # :nodoc:
+        generator.add_text(summary)
+      end
+      # Exact probability of finding values of U lower or equal to sample on U distribution. Use with caution with m*n>100000.
+      # Uses u_sampling_distribution_as62
       def exact_probability
         dist=UMannWhitney.u_sampling_distribution_as62(@n1,@n2)
         sum=0
@@ -151,16 +160,23 @@ Z: #{sprintf("%0.3f",z)} (p: #{sprintf("%0.3f",z_probability)})
         }
         sum
       end
-      # Reference: http://europe.isixsigma.com/library/content/c080806a.asp
+      # Adjunt for ties.
+      # 
+      # Reference: 
+      # * http://europe.isixsigma.com/library/content/c080806a.asp
       def adjust_for_ties(data)
         @t=data.frequencies.find_all{|k,v| v>1}.inject(0) {|a,v|
           a+(v[1]**3-v[1]).quo(12)
         }        
       end
+      
+      private :adjust_for_ties
+      
       # Z value for U, with adjust for ties.
       # For large samples, U is approximately normally distributed. 
       # In that case, you can use z to obtain probabily for U.
-      # Reference: SPSS Manual
+      # Reference: 
+      # * SPSS Manual
       def z
         mu=(@n1*@n2).quo(2)
         if(!@ties)

data/lib/statsample/test.rb

@@ -3,6 +3,8 @@ module Statsample
   
   module Test
     autoload(:UMannWhitney, 'statsample/test/umannwhitney')
+    autoload(:Levene, 'statsample/test/levene')
+
     # Calculate chi square for two Matrix
     class << self
       def chi_square(real,expected)