statsample 0.14.1 → 0.15.0

data/lib/statsample/reliability/scaleanalysis.rb

@@ -18,7 +18,9 @@ module Statsample
         @ds=ds.dup_only_valid
         @k=@ds.fields.size
         @total=@ds.vector_sum
-        @item_mean=@ds.vector_mean.mean
+        @vector_mean=@ds.vector_mean
+        @item_mean=@vector_mean.mean
+        @item_sd=@vector_mean.sd
         @mean=@total.mean
         @median=@total.median
         @skew=@total.skew
@@ -115,8 +117,11 @@ module Statsample
           }
         end
       end
+      # =Adjusted R.P.B. for each item
+      # Adjusted RPB(Point biserial-correlation) for each item
+      #
       def item_total_correlation
-        @ds.fields.inject({}) do |a,v|
+        @itc||=@ds.fields.inject({}) do |a,v|
           vector=@ds[v].clone
           ds2=@ds.clone
           ds2.delete_vector(v)
@@ -125,6 +130,9 @@ module Statsample
           a
         end
       end
+      def mean_rpb
+        item_total_correlation.values.to_scale.mean
+      end
       def item_statistics
           @is||=@ds.fields.inject({}) do |a,v|
             a[v]={:mean=>@ds[v].mean, :sds=>Math::sqrt(@cov_m.variance(v))}
@@ -183,20 +191,24 @@ module Statsample
           t.row [_("Sum median"),   @median]
           t.hr
           t.row [_("Item mean"),    "%0.4f" % @item_mean]
+          t.row [_("Item sd"),    "%0.4f" % @item_sd]
+          t.hr
           t.row [_("Skewness"),     "%0.4f" % @skew]
           t.row [_("Kurtosis"),     "%0.4f" % @kurtosis]
           t.hr
           t.row [_("Cronbach's alpha"), "%0.4f" % @alpha]
           t.row [_("Standarized Cronbach's alpha"), "%0.4f" % @alpha_standarized]
+          t.row [_("Mean rpb"), "%0.4f" % mean_rpb]
+          
           t.row [_("Variances mean"),  "%g" % @variances_mean]
           t.row [_("Covariances mean") , "%g" % @covariances_mean]
           end
-          s.text _("items for obtain alpha(0.8) : %d" % Statsample::Reliability::n_for_desired_alpha(0.8, @variances_mean,@covariances_mean))
-          s.text _("items for obtain alpha(0.9) : %d" % Statsample::Reliability::n_for_desired_alpha(0.9, @variances_mean,@covariances_mean))          
-          itc=item_total_correlation
+          s.text _("Items for obtain alpha(0.8) : %d" % Statsample::Reliability::n_for_desired_reliability(@alpha, 0.8, @ds.fields.size))
+          s.text _("Items for obtain alpha(0.9) : %d" % Statsample::Reliability::n_for_desired_reliability(@alpha, 0.9, @ds.fields.size))          
+          
           sid=stats_if_deleted
           is=item_statistics
-          
+          itc=item_total_correlation
           
           
           s.table(:name=>_("Items report for %s") % @name, :header=>["item","mean","sd", "mean if deleted", "var if deleted", "sd if deleted"," item-total correl.", "alpha if deleted"]) do |t|

data/lib/statsample/srs.rb

@@ -1,6 +1,11 @@
 module Statsample
 	# Several methods to estimate parameters for simple random sampling
+  # == Reference: 
+  # * Cochran, W.(1972). Sampling Techniques [spanish edition].
+  # * http://stattrek.com/Lesson6/SRS.aspx
+  
 	module SRS
+
 		class << self
       ########################
       #
@@ -71,7 +76,6 @@ module Statsample
       # Know proportion, sample without replacement.
       #
       # Sources: 
-      # * http://stattrek.com/Lesson6/SRS.aspx
       # * Cochran(1972)
       def proportion_sd_kp_wor(p, sam, pop)
           fpc(sam,pop)*Math::sqrt(p*(1-p).quo(sam))

data/lib/statsample/test.rb

@@ -32,7 +32,6 @@ module Statsample
     extend self
     # Calculate chi square for two Matrix
     class << self
-      
       def chi_square(observed, expected=nil)
         case observed
           when Vector
@@ -41,7 +40,7 @@ module Statsample
             ChiSquare::WithMatrix.new(observed,expected)
           else
             raise "Not implemented for #{observed.class}"
-          end
+        end
       end
       # Shorthand for Statsample::Test::UMannWhitney.new
       # 

data/lib/statsample/test/f.rb

@@ -41,7 +41,11 @@ module Statsample
         p_using_cdf(Distribution::F.cdf(f, @df_num, @df_den), tails)
       end
       def report_building(builder) #:nodoc:
-        builder.text "%s : F(%d, %d) = %0.4f , p = %0.4f" % [@name, @df_num, @df_den, f, probability]
+        if @df_num.is_a? Integer and @df_den.is_a? Integer
+          builder.text "%s : F(%d, %d) = %0.4f , p = %0.4f" % [@name, @df_num, @df_den, f, probability]
+        else
+          builder.text "%s : F(%0.2f, %0.2f) = %0.4f , p = %0.4f" % [@name, @df_num, @df_den, f, probability]
+        end
       end
     end
   end

data/lib/statsample/test/umannwhitney.rb

@@ -25,7 +25,7 @@ module Statsample
       # Parameters:
       # * <tt>n1</tt>: group 1 size
       # * <tt>n2</tt>: group 2 size 
-      # Reference: 
+      # == 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)
@@ -169,7 +169,7 @@ module Statsample
       end
       # Adjunt for ties.
       # 
-      # Reference: 
+      # == 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|
@@ -182,7 +182,7 @@ module Statsample
       # 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: 
+      # == Reference: 
       # * SPSS Manual
       def z
         mu=(@n1*@n2).quo(2)

data/lib/statsample/vector.rb

@@ -361,7 +361,7 @@ module Statsample
               sum.push(nil)
           end
       }
-      Statsample::Vector.new(sum)
+      Statsample::Vector.new(sum, :scale  )
       else
       raise ArgumentError, "The array/vector parameter should be of the same size of the original vector"
       end
@@ -373,8 +373,7 @@ module Statsample
           else
             nil
           end
-        }
-      )
+        } , :scale)
     else
         raise TypeError,"You should pass a scalar or a array/vector"
     end
@@ -761,7 +760,6 @@ module Statsample
       m||=mean
       @scale_data.inject(0){|a,x| a+(x-m).square}
     end
-    
     # Sum of squared deviation
     def sum_of_squared_deviation
       check_type :scale

data/references.txt

@@ -0,0 +1,22 @@
+References
+* Azen, R. & Budescu, D.V. (2003). The dominance analysis approach for comparing predictors in multiple regression. <em>Psychological Methods, 8</em>(2), 129-148.
+* Azen, R. & Budescu, D.V. (2006). Comparing predictors in Multivariate Regression Models: An extension of Dominance Analysis. <em>Journal of Educational and Behavioral Statistics, 31</em>(2), 157-180.
+* Budescu, D. V. (1993). Dominance analysis: a new approach to the problem of relative importance of predictors in multiple regression. <em>Psychological Bulletin, 114</em>, 542-551.
+* Cochran(1972)
+* Cohen et al. (2003). Applied Multiple Reggression / Correlation Analysis for the Behavioral Sciences
+* 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
+* Dziuban, C., & Shirkey E. (1974). When is a correlation matrix appropriate for factor analysis? Some decision rules. Psychological Bulletin, 81(6), 358-361.
+* Hayton, J., Allen, D. & Scarpello, V.(2004). Factor Retention Decisions in Exploratory Factor Analysis: a Tutorial on Parallel Analysis. <i>Organizational Research Methods, 7</i> (2), 191-205.
+* Lin, J. (2007). VARIMAX_K58 [Source code]. [http://www.johnny-lin.com/idl_code/varimax_k58.pro]
+* Liu, O., & Rijmen, F. (2008). A modified procedure for parallel analysis of ordered categorical data. Behavior Research Methods, 40(2), 556-562.
+* McGraw, K. & Wong, S.P. (1996). Forming Inferences About Some Intraclass Correlation Coefficients. Psychological methods, 1(1), 30-46.
+* O'Connor, B. (2000). SPSS and SAS programs for determining the number of components using parallel analysis and Velicer's MAP test. Behavior Research Methods, Instruments, & Computers, 32(3), 396-402.
+* SPSS Manual
+* Shrout,P. & Fleiss, J. (1979). Intraclass Correlation: Uses in assessing rater reliability. Psychological Bulletin, 86(2), 420-428
+* Smith, L. (2002). A tutorial on Principal Component Analysis. Available on http://courses.eas.ualberta.ca/eas570/pca_tutorial.pdf 
+* http://en.wikipedia.org/wiki/Welch-Satterthwaite_equation
+* http://europe.isixsigma.com/library/content/c080806a.asp
+* http://snippets.dzone.com/posts/show/4666
+* http://stattrek.com/Lesson6/SRS.aspx
+* http://www.cut-the-knot.org/do_you_know/AllPerm.shtml
+* http://www.gnu.org/software/gsl/manual/html_node/The-histogram-struct.html

data/test/test_bivariate.rb

@@ -1,12 +1,12 @@
 require(File.dirname(__FILE__)+'/helpers_tests.rb')
 
 class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
-  def test_sum_of_squares
+  should "method sum of squares should be correct" do
     v1=[1,2,3,4,5,6].to_vector(:scale)
     v2=[6,2,4,10,12,8].to_vector(:scale)
     assert_equal(23.0, Statsample::Bivariate.sum_of_squares(v1,v2))
   end
-  def test_covariance
+  should "return same covariance with ruby and gls implementation" do
     if Statsample.has_gsl?
       v1=20.times.collect {|a| rand()}.to_scale
       v2=20.times.collect {|a| rand()}.to_scale
@@ -17,7 +17,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
 
   end
 
-  def test_gsl_pearson
+  should "return same correlation with ruby and gls implementation" do
     if Statsample.has_gsl?
       v1=20.times.collect {|a| rand()}.to_scale
       v2=20.times.collect {|a| rand()}.to_scale
@@ -27,7 +27,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
       skip "Not tested gsl versus ruby correlation (needs GSL)"
     end
   end
-  def test_pearson
+  should "return correct pearson correlation" do
     v1=[6,5,4,7,8,4,3,2].to_vector(:scale)
     v2=[2,3,7,8,6,4,3,2].to_vector(:scale)
     assert_in_delta(0.525,Statsample::Bivariate.pearson(v1,v2), 0.001)
@@ -40,7 +40,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
     v3a,v4a=Statsample.only_valid v3, v4
     assert_in_delta(0.525, Statsample::Bivariate.pearson_slow(v3a,v4a),0.001)
   end
-  def test_bivariate_pearson
+  should "return correct values for t_pearson and prop_pearson" do
     v1=[6,5,4,7,8,4,3,2].to_vector(:scale)
     v2=[2,3,7,8,6,4,3,2].to_vector(:scale)
     r=Statsample::Bivariate::Pearson.new(v1,v2)
@@ -49,7 +49,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
     assert_in_delta(Statsample::Bivariate.prop_pearson(r.t,8,:both), r.probability, 0.001)
     assert(r.summary.size>0)
   end
-  def test_matrix_correlation
+  should "return correct correlation_matrix" do
     v1=[6,5,4,7,8,4,3,2].to_vector(:scale)
     v2=[2,3,7,8,6,4,3,2].to_vector(:scale)
     v3=[6,2,  1000,1000,5,4,7,8].to_vector(:scale)
@@ -69,7 +69,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
     end
     #assert_equal(expected,obt)
   end
-  def test_prop_pearson
+  should "return correct value for prop pearson" do
     assert_in_delta(0.42, Statsample::Bivariate.prop_pearson(Statsample::Bivariate.t_r(0.084,94), 94),0.01)
     assert_in_delta(0.65, Statsample::Bivariate.prop_pearson(Statsample::Bivariate.t_r(0.046,95), 95),0.01)
     r=0.9
@@ -87,13 +87,13 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
     assert(Statsample::Bivariate.prop_pearson(t,n,:left)<0.05)
   end
 
-  def test_spearman
+  should "return correct value for Spearman's rho" do
     v1=[86,97,99,100,101,103,106,110,112,113].to_vector(:scale)
     v2=[0,20,28,27,50,29,7,17,6,12].to_vector(:scale)
     assert_in_delta(-0.175758,Statsample::Bivariate.spearman(v1,v2),0.0001)
 
   end
-  def test_point_biserial
+  should "return correct value for point_biserial correlation" do
     c=[1,3,5,6,7,100,200,300,400,300].to_vector(:scale)
     d=[1,1,1,1,1,0,0,0,0,0].to_vector(:scale)
     assert_raises TypeError do
@@ -101,7 +101,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
     end
     assert_in_delta(Statsample::Bivariate.point_biserial(d,c), Statsample::Bivariate.pearson(d,c), 0.0001)
   end
-  def test_tau
+  should "return correct value for tau_a and tau_b" do
     v1=[1,2,3,4,5,6,7,8,9,10,11].to_vector(:ordinal)
     v2=[1,3,4,5,7,8,2,9,10,6,11].to_vector(:ordinal)
     assert_in_delta(0.6727,Statsample::Bivariate.tau_a(v1,v2),0.001)
@@ -110,7 +110,7 @@ class StatsampleBivariateTestCase < MiniTest::Unit::TestCase
     v2=[11,4,4,2,0,0,0,0,0,0,4,0,4,0,0,0,0,4,0,0,0,0,0].to_vector(:ordinal)
     assert_in_delta(-0.376201540231705, Statsample::Bivariate.tau_b(Statsample::Crosstab.new(v1,v2).to_matrix),0.001)
   end
-  def test_gamma
+  should "return correct value for gamma correlation" do
     m=Matrix[[10,5,2],[10,15,20]]
     assert_in_delta(0.636,Statsample::Bivariate.gamma(m),0.001)
     m2=Matrix[[15,12,6,5],[12,8,10,8],[4,6,9,10]]

data/test/test_reliability.rb

@@ -1,6 +1,16 @@
 require(File.dirname(__FILE__)+'/helpers_tests.rb')
 class StatsampleReliabilityTestCase < MiniTest::Unit::TestCase
   context Statsample::Reliability do
+    should "return correct r according to Spearman-Brown prophecy" do
+      r=0.6849
+      n=62.quo(15)
+      assert_in_delta(0.9, Statsample::Reliability.sbp(r,n), 0.001)
+    end
+    should "return correct n for desired realiability" do
+        r=0.6849
+        r_d=0.9
+        assert_in_delta(62, Statsample::Reliability.n_for_desired_reliability(r, r_d, 15),0.5)
+      end
     context "Cronbach's alpha" do 
       setup do
         @samples=40
@@ -44,6 +54,7 @@ class StatsampleReliabilityTestCase < MiniTest::Unit::TestCase
         #p n_obtained
         assert_in_delta(Statsample::Reliability.cronbach_alpha_from_n_s2_cov(n_obtained, vm,cm) ,@a,0.001) 
       end
+      
       should "standarized alpha will be equal to sum of matrix covariance less the individual variances on standarized values" do
         total_sum=@cme.total_sum
         ind_var=@dse.fields.inject(0) {|ac,v| ac+@dse[v].variance}

data/test/test_reliability_icc.rb

@@ -0,0 +1,188 @@
+$reliability_icc=nil
+require(File.dirname(__FILE__)+'/helpers_tests.rb')
+class StatsampleReliabilityIccTestCase < MiniTest::Unit::TestCase
+  context Statsample::Reliability::ICC do
+    setup do
+      a=[9,6,8,7,10,6].to_scale
+      b=[2,1,4,1,5,2].to_scale
+      c=[5,3,6,2,6,4].to_scale
+      d=[8,2,8,6,9,7].to_scale
+      @ds={'a'=>a,'b'=>b,'c'=>c,'d'=>d}.to_dataset
+      @icc=Statsample::Reliability::ICC.new(@ds)
+    end
+    should "basic method be correct" do
+      assert_equal(6,@icc.n)
+      assert_equal(4,@icc.k)
+    end
+    should "total mean be correct" do
+      assert_in_delta(5.291, @icc.total_mean, 0.001)
+    end
+    should "df methods be correct" do
+      assert_equal(5,  @icc.df_bt)
+      assert_equal(18, @icc.df_wt)
+      assert_equal(3,  @icc.df_bj)
+      assert_equal(15, @icc.df_residual)
+    end
+    should "ms between targets be correct" do
+      assert_in_delta(11.24, @icc.ms_bt, 0.01)
+    end
+    should "ms within targets be correct" do
+      assert_in_delta(6.26,  @icc.ms_wt, 0.01)
+    end
+    should "ms between judges be correct" do 
+      assert_in_delta(32.49, @icc.ms_bj, 0.01)
+    end
+    should "ms residual be correct" do
+      assert_in_delta(1.02,  @icc.ms_residual, 0.01)
+    end
+    context "with McGraw and Wong denominations," do 
+      
+    end
+    context "with Shrout & Fleiss denominations, " do 
+      should "icc(1,1) method be correct" do
+        assert_in_delta(0.17, @icc.icc_1_1, 0.01)
+      end
+      # Verified on SPSS and R
+      should "icc(2,1) method be correct" do    
+        assert_in_delta(0.29, @icc.icc_2_1, 0.01)
+      end
+      should "icc(3,1) method be correct" do
+        assert_in_delta(0.71, @icc.icc_3_1, 0.01)
+      end
+      should "icc(1,k) method be correct" do
+        assert_in_delta(0.44, @icc.icc_1_k, 0.01)
+      end
+      # Verified on SPSS and R
+      should "icc(2,k) method be correct" do
+        assert_in_delta(0.62, @icc.icc_2_k, 0.01)
+      end 
+      should "icc(3,k) method be correct" do
+        assert_in_delta(0.91, @icc.icc_3_k, 0.01)
+      end
+      
+      should "icc(1,1) F be correct" do
+        assert_in_delta(1.795, @icc.icc_1_f.f)
+      end
+      should "icc(1,1) confidence interval should be correct" do
+        assert_in_delta(-0.133, @icc.icc_1_1_ci[0], 0.001)
+        assert_in_delta(0.723, @icc.icc_1_1_ci[1], 0.001)
+      end
+      should "icc(1,k) confidence interval should be correct" do
+        assert_in_delta(-0.884, @icc.icc_1_k_ci[0], 0.001)
+        assert_in_delta(0.912, @icc.icc_1_k_ci[1], 0.001)
+      end
+      
+      should "icc(2,1) F be correct" do
+        assert_in_delta(11.027, @icc.icc_2_f.f)
+      end
+      should "icc(2,1) confidence interval should be correct" do
+        #skip("Not yet operational")
+        assert_in_delta(0.019, @icc.icc_2_1_ci[0], 0.001)
+        assert_in_delta(0.761, @icc.icc_2_1_ci[1], 0.001)
+      end
+      
+      # Verified on SPSS and R      
+      should "icc(2,k) confidence interval should be correct" do
+        #skip("Not yet operational")
+        #p @icc.icc_2_k_ci
+        assert_in_delta(0.039, @icc.icc_2_k_ci[0], 0.001)
+        assert_in_delta(0.929, @icc.icc_2_k_ci[1], 0.001)
+
+      end
+      #should "Shrout icc(2,k) and McGraw icc(a,k) ci be equal" do
+      #  assert_in_delta(@icc.icc_2_k_ci_shrout[0], @icc.icc_2_k_ci_mcgraw[0], 10e-5)
+      #end
+      
+      should "icc(3,1) F be correct" do
+        assert_in_delta(11.027, @icc.icc_3_f.f)
+      end
+      
+      should "icc(3,1) confidence interval should be correct" do
+        assert_in_delta(0.342, @icc.icc_3_1_ci[0], 0.001)
+        assert_in_delta(0.946, @icc.icc_3_1_ci[1], 0.001)
+      end
+      should "icc(3,k) confidence interval should be correct" do
+        assert_in_delta(0.676, @icc.icc_3_k_ci[0], 0.001)
+        assert_in_delta(0.986, @icc.icc_3_k_ci[1], 0.001)
+      end
+    end
+    
+    begin
+      require 'rserve'
+      require 'statsample/rserve_extension'
+      context "McGraw and Wong" do
+        setup do
+          if($reliability_icc.nil?)
+            size=100
+            a=size.times.map {rand(10)}.to_scale
+            b=a.recode{|i|i+rand(4)-2}
+            c=a.recode{|i|i+rand(4)-2}
+            d=a.recode{|i|i+rand(4)-2}
+            @ds={'a'=>a,'b'=>b,'c'=>c,'d'=>d}.to_dataset
+            
+            
+            
+            @icc=Statsample::Reliability::ICC.new(@ds)
+            @r=Rserve::Connection.new
+            @r.assign('ds',@ds)
+            @r.void_eval("library(irr);
+              iccs=list(
+              icc_1=icc(ds,'o','c','s'),
+              icc_k=icc(ds,'o','c','a'),
+              icc_c_1=icc(ds,'t','c','s'),
+              icc_c_k=icc(ds,'t','c','a'),
+              icc_a_1=icc(ds,'t','a','s'),
+              icc_a_k=icc(ds,'t','a','a'))
+              ")
+            @iccs=@r.eval('iccs').to_ruby
+            $reliability_icc={ :icc=>@icc, :iccs=>@iccs
+            }
+          end
+          @icc=$reliability_icc[:icc]
+          @iccs=$reliability_icc[:iccs]
+          
+        end
+        [:icc_1, :icc_k, :icc_c_1, :icc_c_k, :icc_a_1, :icc_a_k].each do |t|
+          context "ICC Type #{t}" do
+            should "value be correct" do
+              @icc.type=t
+              @r_icc=@iccs[t.to_s]
+              assert_in_delta(@r_icc['value'],@icc.r)
+            end
+            should "fvalue be correct" do
+              @icc.type=t
+              @r_icc=@iccs[t.to_s]
+              assert_in_delta(@r_icc['Fvalue'],@icc.f.f)
+            end
+            should "num df be correct" do
+              @icc.type=t
+              @r_icc=@iccs[t.to_s]
+              assert_in_delta(@r_icc['df1'],@icc.f.df_num)
+            end
+            should "den df be correct" do
+              @icc.type=t
+              @r_icc=@iccs[t.to_s]
+              assert_in_delta(@r_icc['df2'],@icc.f.df_den)
+            end
+
+            should "f probability be correct" do
+              @icc.type=t
+              @r_icc=@iccs[t.to_s]
+              assert_in_delta(@r_icc['p.value'],@icc.f.probability)
+            end
+            should "bounds be equal" do
+              @icc.type=t
+              @r_icc=@iccs[t.to_s]
+              assert_in_delta(@r_icc['lbound'],@icc.lbound)
+              assert_in_delta(@r_icc['ubound'],@icc.ubound)
+              
+            end
+          end
+        end
+      end
+    rescue
+      puts "requires rserve"
+    end
+    
+  end
+end