num4regana 0.0.1-java → 0.0.2-java

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 616a5ce4aa14748f31750e7b1c2e69ca9af26295e7a173a8454ff9b3bc83cbfe
-  data.tar.gz: 74b2370bb501a69ef6546840d3ed0b61315e13ef35f584d5dbe3f2f38725cbb2
+  metadata.gz: ad6fa72619eb7a02bdc561494b9aae197c37b6a8558fd21106bdf15210a9af81
+  data.tar.gz: c7a55ac44970c164dc7379a11955f29416392ceba4605118f8bcbba459724e94
 SHA512:
-  metadata.gz: 7133ab852f61abbcb0e93bbc1be555bcfe16dfab0751ff1026be21dc8a46d829b3d03ba94313a7a196a0f385b7dbe01a6cb16d4678337bb063662e3faac88b43
-  data.tar.gz: a9851b753e8bf686a1a5bccc5dc392dc229822e51e47a3b1b82c8f1f76dbaf8629c04ecc825679aee250ff1225a7cf9ab6ed2f774f61e927f32fcb58ba83e7ee
+  metadata.gz: dd9e09e45d35384a3bfa9eb50a0132f4b3b3e970c2f97d241264a18a08986c3a46753d2292caf10c0ae1fcab8ec4f45f6456c97fe14df2ee9334168cd429a60a
+  data.tar.gz: 1cf8a55a5444c5e1e1a151adbb4d142bceaa9344cde28a859848b5642ce36660b6b919410907f8964105ec35df991591670ba8722658a708822e0840d790ce23

data/CHANGELOG.md

@@ -2,6 +2,11 @@
 
 ## Unreleased
 
+## [0.0.2] - 2024-08-08
+
+### fix
+- add function of equal variances in OLSMultRegAnaLib
+
 ## [0.0.1] - 2024-06-27
 
 ### Fixed

data/ext/num4regana/MultRegAna.java

@@ -0,0 +1,206 @@
+import org.apache.commons.math3.stat.regression.OLSMultipleLinearRegression;
+import org.apache.commons.math3.stat.regression.GLSMultipleLinearRegression;
+import org.apache.commons.math3.stat.correlation.Covariance;
+import org.apache.commons.math3.stat.descriptive.DescriptiveStatistics;
+import java.util.Arrays;
+import org.apache.commons.math3.distribution.ChiSquaredDistribution;
+
+public class MultRegAna {
+    private final double A = 0.05;
+    private static MultRegAna regana = new MultRegAna();
+    public static MultRegAna getInstance() {
+        return regana;
+    }
+    public LineReg lineRegAna(double[] yi, double xij[][]) {
+        double[][] data = createData(yi, xij);
+        LineRegAna line = createLineRegAna(data);
+
+        return line.lineRegAna(yi, xij);
+    }
+    public double getR2(double[] yi, double xij[][]) {
+        double[][] data = createData(yi, xij);
+        LineRegAna line = createLineRegAna(data);
+
+        return line.getR2(yi, xij);
+    }
+    public double getAdjR2(double[] yi, double xij[][]) {
+        double[][] data = createData(yi, xij);
+        LineRegAna line = createLineRegAna(data);
+
+        return line.getAdjR2(yi, xij);
+    }
+    private double[][] createData(double[] yi, double xij[][]) {
+        double[][] data = new double[yi.length][1 + xij[0].length];
+
+        for (int i = 0; i < yi.length; i++) {
+            data[i][0] = yi[i];
+            System.arraycopy(xij[i], 0, data[i], 1, xij[0].length);
+        }
+        return data;
+    }
+    private LineRegAna createLineRegAna(double data[][]) {
+        // 等分散性の検定
+        if (false == bartletTest(data)) {  // 等分散性
+            return new OLSMultRegAna();
+        }
+        else {                             // 
+            return new GLSMultRegAna(data);
+        }
+    }
+    private boolean bartletTest(double data[][]) {
+        OneWayAnovaTest anova = new BartletTest();
+        double statistic = anova.calcTestStatistic(data);
+
+        return anova.execute_test(statistic, A);
+    }
+    /*********************************/
+    /* interface define              */
+    /*********************************/
+    private interface LineRegAna {
+        // 最小２乗法
+        LineReg lineRegAna(double[] yi, double xij[][]);
+        // 決定係数取得
+        double getR2(double[] yi, double xij[][]);
+        // 自由度調整済み決定係数
+        double getAdjR2(double[] yi, double xij[][]);
+    }
+    private interface OneWayAnovaTest {
+        double calcTestStatistic(double[][] xi);
+        boolean execute_test(double statistic, double a);
+    }
+    /*********************************/
+    /* class define                  */
+    /*********************************/
+    public class LineReg {
+        private double a   = 0.0;
+        private double[] b = null;
+        public LineReg(double[] b) {
+            this.a = b[0];
+            this.b = new double[b.length - 1];
+            for (int i = 0; i < this.b.length; i++) {
+                this.b[i] = b[i + 1];
+            }
+        }
+        public double getIntercept() {
+            return a;
+        }
+        public double[] getSlope() {
+            return b;
+        }
+    }
+    // 等分散性検定
+    private class BartletTest implements OneWayAnovaTest {
+        private int n = 0;
+        public double calcTestStatistic(double[][] xi) {
+            n = xi.length;
+            double ln2L = logL(xi);
+
+            return calcB(ln2L, xi);
+        }
+        private double logL(double[][] xi) {
+            double[] si = new double[n];
+            DescriptiveStatistics stat = new DescriptiveStatistics();
+            double nisi2 = 0.0;    // (Ni - 1)*si^2の合計
+            double nilogsi2 = 0.0; // (Ni - 1)*log(si^2)の合計
+            int sumN = 0;
+
+            for(int i = 0; i < n; i++) {
+                Arrays.stream(xi[i]).forEach(stat::addValue);
+                sumN += stat.getN();
+                si[i] = stat.getVariance();
+                nisi2 += (stat.getN() - 1) * si[i];
+                nilogsi2 += (stat.getN() - 1) * Math.log(si[i]);
+                stat.clear();
+            }
+            double sumNin = sumN - n;
+            return sumNin * (Math.log(nisi2 / sumNin) - nilogsi2 / sumNin);
+        }
+        private double calcB(double ln2L, double[][] xi) {
+            double invSumN = 0.0;
+            int sumN = 0;
+            DescriptiveStatistics stat = new DescriptiveStatistics();
+
+            for(int i = 0; i < n; i++) {
+                Arrays.stream(xi[i]).forEach(stat::addValue);
+                invSumN += 1.0 / (stat.getN() - 1.0);
+                sumN += stat.getN();
+                stat.clear();
+            }
+            double deno = 1 + 1.0 / (3 * (n - 1))
+                        * (invSumN - 1.0 / (sumN - n));
+            return ln2L / deno;
+        }
+        public boolean execute_test(double statistic, double a) {
+            ChiSquaredDistribution chi2Dist = new ChiSquaredDistribution(n - 1);
+            double r_val = chi2Dist.inverseCumulativeProbability(1.0 - a);
+
+            return (r_val < statistic) ? true : false;
+        }
+    }
+
+    // 最小２乗法
+    private class OLSMultRegAna implements LineRegAna {
+        private OLSMultipleLinearRegression regression = null;
+        public OLSMultRegAna() {
+            regression = new OLSMultipleLinearRegression();
+        }
+        public LineReg lineRegAna(double[] yi, double xij[][]) {
+            regression.newSampleData(yi, xij);
+
+            double[] beta = regression.estimateRegressionParameters();
+
+            return new LineReg(beta);
+        }
+        // 決定係数取得
+        public double getR2(double[] yi, double xij[][]) {
+            regression.newSampleData(yi, xij);
+            return regression.calculateRSquared();
+        }
+        // 自由度調整済み決定係数
+        public double getAdjR2(double[] yi, double xij[][]) {
+            regression.newSampleData(yi, xij);
+            return regression.calculateAdjustedRSquared();
+        }
+        
+    }
+    // 一般化最小２乗法
+    private class GLSMultRegAna implements LineRegAna {
+        private GLSMultipleLinearRegression regression = null;
+        private double[][] data = null;
+        public GLSMultRegAna(double data[][]) {
+            regression = new GLSMultipleLinearRegression();
+            this.data = data;
+        }
+        public LineReg lineRegAna(double[] yi, double xij[][]) {
+            double[][] omega = calcCovatrianceMatrix();
+            regression.newSampleData(yi, xij, omega);
+
+            double[] beta = regression.estimateRegressionParameters();
+            return new LineReg(beta);
+        }
+        // 決定係数取得
+        public double getR2(double[] yi, double xij[][]) {
+            return 0.0;
+        }
+        // 自由度調整済み決定係数
+        public double getAdjR2(double[] yi, double xij[][]) {
+            return 0.0;
+        }
+        private double[][] calcCovatrianceMatrix() {
+            Covariance corel = new Covariance();
+            double[][] omega = new double[data.length][data.length];
+
+            for(int i = 0; i < data.length; i++) {
+                for(int j = 0; j < data.length; j++) {
+                    double[] xArray = data[i];
+                    double[] yArray = data[j];
+
+                    omega[i][j] = corel.covariance(xArray, yArray);
+                }
+            }
+            return omega;
+        }
+    }
+            
+}
+

data/lib/num4regana.rb

@@ -3,7 +3,7 @@ require 'num4regana.jar'
 require 'commons-math3-3.6.1.jar'
 
 java_import 'SmplRegAna'
-java_import 'OLSMultRegAna'
+java_import 'MultRegAna'
 # 回帰分析
 #  (Apache commoms math3使用)
 module Num4RegAnaLib
@@ -67,10 +67,10 @@ module Num4RegAnaLib
             return @regana.getR(yi.to_java(Java::double), xi.to_java(Java::double))
         end
     end
-    # 重回帰分析(最小２乗法)
+    # 重回帰分析(最小２乗法:等分散性checkあり)
     class OLSMultRegAnaLib
         def initialize
-            @regana = OLSMultRegAna.getInstance()
+            @multana = MultRegAna.getInstance()
         end
         # 重回帰分析
         #
@@ -100,10 +100,11 @@ module Num4RegAnaLib
         #        "slope":      [3.47, 0.53],     # 回帰係数
         #     }
         def line_reg_ana(yi, xij)
-            ret = @regana.lineRegAna(yi.to_java(Java::double), xij.to_java(Java::double[]))
+            multRet = @multana.lineRegAna(yi.to_java(Java::double), xij.to_java(Java::double[]))
+
             retRb = {
-                "intercept":  ret.getIntercept(), # 定数項
-                "slope":      ret.getSlope().to_a,     # 回帰係数
+                "intercept":  multRet.getIntercept(), # 定数項
+                "slope":      multRet.getSlope().to_a,     # 回帰係数
             }
             return retRb
         end
@@ -131,7 +132,7 @@ module Num4RegAnaLib
         #   regana.getr2(yi, xi)
         #   => 0.858
         def getr2(yi, xij)
-            return @regana.getR2(yi.to_java(Java::double), xij.to_java(Java::double[]))
+            return @multana.getR2(yi.to_java(Java::double), xij.to_java(Java::double[]))
         end
         # 自由度調整済み決定係数
         #
@@ -157,7 +158,7 @@ module Num4RegAnaLib
         #   regana.getadjr2(yi, xij)
         #   => 0.8176
         def getadjr2(yi, xij)
-            return @regana.getAdjR2(yi.to_java(Java::double), xij.to_java(Java::double[]))
+            return @multana.getAdjR2(yi.to_java(Java::double), xij.to_java(Java::double[]))
         end
 
     end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: num4regana
 version: !ruby/object:Gem::Version
-  version: 0.0.1
+  version: 0.0.2
 platform: java
 authors:
 - siranovel
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2024-06-27 00:00:00.000000000 Z
+date: 2024-08-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rake
@@ -61,7 +61,7 @@ files:
 - Gemfile
 - LICENSE
 - Rakefile
-- ext/num4regana/OLSMultRegAna.java
+- ext/num4regana/MultRegAna.java
 - ext/num4regana/SmplRegAna.java
 - lib/commons-math3-3.6.1.jar
 - lib/num4regana.rb

data/ext/num4regana/OLSMultRegAna.java

@@ -1,75 +0,0 @@
-import org.apache.commons.math3.stat.regression.OLSMultipleLinearRegression;
-
-public class OLSMultRegAna {
-    private static OLSMultRegAna regana = new OLSMultRegAna();
-    public static OLSMultRegAna getInstance() {
-        return regana;
-    }
-    public LineReg lineRegAna(double[] yi, double xij[][]) {
-        LineRegAna line = new LineRegAna();
-
-        return line.lineRegAna(yi, xij);
-    }
-    public double getR2(double[] yi, double xij[][]) {
-        LineRegAna line = new LineRegAna();
-
-        return line.getR2(yi, xij);
-    }
-    public double getAdjR2(double[] yi, double xij[][]) {
-        LineRegAna line = new LineRegAna();
-
-        return line.getAdjR2(yi, xij);
-    }
-    /*********************************/
-    /* interface define              */
-    /*********************************/
-    /*********************************/
-    /* class define                  */
-    /*********************************/
-    public class LineReg {
-        private double a   = 0.0;
-        private double[] b = null;
-        public LineReg(double[] b) {
-            this.a = b[0];
-            this.b = new double[b.length - 1];
-            for (int i = 0; i < this.b.length; i++) {
-                this.b[i] = b[i + 1];
-            }
-        }
-        public double getIntercept() {
-            return a;
-        }
-        public double[] getSlope() {
-            return b;
-        }
-    }
-    private class LineRegAna {
-        // 最小２乗法
-        public LineReg lineRegAna(double[] yi, double xij[][]) {
-            OLSMultipleLinearRegression regression = new OLSMultipleLinearRegression();
-
-            regression.newSampleData(yi, xij);
-
-            double[] beta = regression.estimateRegressionParameters();
-
-            LineReg ret = new LineReg(beta);
-            return ret;
-        }
-        // 決定係数取得
-        public double getR2(double[] yi, double xij[][]) {
-            OLSMultipleLinearRegression regression = new OLSMultipleLinearRegression();
-
-            regression.newSampleData(yi, xij);
-            return regression.calculateRSquared();
-        }
-        // 自由度調整済み決定係数
-        public double getAdjR2(double[] yi, double xij[][]) {
-            OLSMultipleLinearRegression regression = new OLSMultipleLinearRegression();
-
-            regression.newSampleData(yi, xij);
-            return regression.calculateAdjustedRSquared();
-        }
-        
-    }
-}
-