Working on documentation.

Release notes.txt

@@ -19,20 +19,25 @@ Version updates and fixes:
   - Improving sample generation for Poisson and Rayleigh distributions;
   - Updating all univariate distributions to support sample generation;
   - Making sure all probability distributions implement IFormattable;
+  - Adding Generalized Beta distribution with PERT estimation;
+  - Adding sample generation in Beta and Generalized Beta distributions;
+  - Adding estimation using the Method-of-moments and Maximum Likelihood;
+  - Adding support for weighted samples in LogisticRegressionAnalysis;
+  - Adding a named constructor to create an Analysis from summary data;
   - Adding all missing Shapiro-Wilk distribution's methods;
   - Adding a common interface for radial basis function kernels;
   - Adding a new generic Gaussian kernel for creating composite kernels;
   - Adding a Windowing filter in the Statistics filters namespace.
 
 * Accord.Math
-  - Updating Augmented Lagragian to detect situations where the 
-    inner optimization algorithm has diverged more accurately;
+  - Updating Augmented Lagragian to detect more accurately 
+    when the inner optimization algorithm has diverged;
   - Adding a new Fast Fourier Transform (FFT) implementation for general 
     matrices and vectors whose dimensions are not necessarily powers of 2.
 
 * Accord.MachineLearning
-  - Tree inducing algorithms (ID3 and C4.5) can now reuse decision 
-    variables multiple times when creating a decision tree;
+  - Updating tree inducing algorithms (ID3 and C4.5) se they can reuse
+    decision variables multiple times when creating a decision tree;
   - Correcting Levenberg-Marquardt's chain-rule Jacobian calculation 
     when there are many output neurons in the learned neural network;
   - Updating the way SVM learning algorithms detect whether a machine is linear or not;

Sources/Accord.Statistics/Analysis/LogisticRegressionAnalysis.cs

@@ -62,6 +62,12 @@ namespace Accord.Statistics.Analysis
     /// </remarks>
     /// 
     /// <example>
+    /// <para>
+    ///   The following example shows to create a Logistic regresion analysis using a full
+    ///   dataset composed of input vectors and a binary output vector. Each input vector
+    ///   has an associated label (1 or 0) in the output vector, where 1 represents a positive
+    ///   label (yes, or true) and 0 represents a negative label (no, or false).</para>
+    ///   
     /// <code>
     /// // Suppose we have the following data about some patients.
     /// // The first variable is continuous and represent patient
@@ -125,6 +131,62 @@ namespace Accord.Statistics.Analysis
     ///             
     /// // For those inputs, the answer probability is approximately 75%.
     /// </code>
+    /// 
+    /// <para>
+    ///   The analysis can also be created from data given in a summary form. Instead of having
+    ///   one input vector associated with one positive or negative label, each input vector is
+    ///   associated with the proportion of positive to negative labels in the original dataset.
+    /// </para>
+    /// 
+    /// <code>
+    /// // Suppose we have a (fictitious) data set about patients who 
+    /// // underwent cardiac surgery. The first column gives the number
+    /// // of arterial bypasses performed during the surgery. The second
+    /// // column gives the number of patients whose surgery went well,
+    /// // while the third column gives the number of patients who had
+    /// // at least one complication during the surgery.
+    /// // 
+    /// int[,] data =
+    /// {
+    ///     // # of stents       success     complications
+    ///     {       1,             140,           45       },
+    ///     {       2,             130,           60       },
+    ///     {       3,             150,           31       },
+    ///     {       4,              96,           65       }
+    /// };
+    /// 
+    /// 
+    /// double[][] inputs = data.GetColumn(0).ToDouble().ToArray();
+    /// 
+    /// int[] positive = data.GetColumn(1);
+    /// int[] negative = data.GetColumn(2);
+    /// 
+    /// // Create a new Logistic Regression Analysis from the summary data
+    /// var regression = LogisticRegressionAnalysis.FromSummary(inputs, positive, negative);
+    /// 
+    /// regression.Compute(); // compute the analysis.
+    /// 
+    /// // Now we can show a summary of the analysis
+    /// DataGridBox.Show(regression.Coefficients);
+    /// 
+    /// 
+    /// // We can also investigate all parameters individually. For
+    /// // example the coefficients values will be available at the
+    /// // vector
+    /// 
+    /// double[] coef = regression.CoefficientValues;
+    /// 
+    /// // The first value refers to the model's intercept term. We
+    /// // can also retrieve the odds ratios and standard errors:
+    /// 
+    /// double[] odds = regression.OddsRatios;
+    /// double[] stde = regression.StandardErrors;
+    /// 
+    /// 
+    /// // Finally, we can use it to estimate risk for a new patient
+    /// double y = regression.Regression.Compute(new double[] { 4 });
+    /// </code>
+    ///   
     /// </example>
     /// 
     [Serializable]

Unit Tests/Accord.Tests.Statistics/Analysis/LogisticRegressionAnalysisTest.cs

@@ -224,30 +224,79 @@ public void ComputeTest4()
         [TestMethod()]
         public void FromSummaryTest1()
         {
+            // Suppose we have a (fictitious) data set about patients who 
+            // underwent cardiac surgery. The first column gives the number
+            // of arterial bypasses performed during the surgery. The second
+            // column gives the number of patients whose surgery went well,
+            // while the third column gives the number of patients who had
+            // at least one complication during the surgery.
+            // 
             int[,] data =
             {
-                { 1, 140, 45 },
-                { 2, 130, 60 },
-                { 3, 150, 31 },
-                { 4,  96, 65 }
+                // # of stents       success     complications
+                {       1,             140,           45       },
+                {       2,             130,           60       },
+                {       3,             150,           31       },
+                {       4,              96,           65       }
             };
 
 
+            double[][] inputs = data.GetColumn(0).ToDouble().ToArray();
+
+            int[] positive = data.GetColumn(1);
+            int[] negative = data.GetColumn(2);
+
+            // Create a new Logistic Regression Analysis from the summary data
+            var regression = LogisticRegressionAnalysis.FromSummary(inputs, positive, negative);
+
+            regression.Compute(); // compute the analysis.
+
+            // Now we can show a summary of the analysis
+            // DataGridBox.Show(regression.Coefficients);
+
+
+            // We can also investigate all parameters individually. For
+            // example the coefficients values will be available at the
+            // vector
+
+            double[] coef = regression.CoefficientValues;
+
+            // The first value refers to the model's intercept term. We
+            // can also retrieve the odds ratios and standard errors:
+
+            double[] odds = regression.OddsRatios;
+            double[] stde = regression.StandardErrors;
+
+
+            // Finally, we can use it to estimate risk for a new patient
+            double y = regression.Regression.Compute(new double[] { 4 });
+
+
+
+
+            Assert.AreEqual(3.7586367581050162, odds[0], 1e-8);
+            Assert.AreEqual(0.85772731075090014, odds[1], 1e-8);
+            Assert.IsFalse(odds.HasNaN());
+
+            Assert.AreEqual(0.20884336554629004, stde[0], 1e-8);
+            Assert.AreEqual(0.075837785246620285, stde[1], 1e-8);
+            Assert.IsFalse(stde.HasNaN());
+
+            Assert.AreEqual(0.67044096045332713, y, 1e-8);
+            Assert.IsFalse(Double.IsNaN(y));
+
             LogisticRegressionAnalysis expected;
-            LogisticRegressionAnalysis actual;
 
 
             {
                 int[] qtr = data.GetColumn(0);
-                int[] positive = data.GetColumn(1);
-                int[] negative = data.GetColumn(2);
 
                 var expanded = Accord.Statistics.Tools.Expand(qtr, positive, negative);
 
-                double[][] inputs = expanded.GetColumn(0).ToDouble().ToArray();
+                double[][] inp = expanded.GetColumn(0).ToDouble().ToArray();
                 double[] outputs = expanded.GetColumn(1).ToDouble();
 
-                expected = new LogisticRegressionAnalysis(inputs, outputs);
+                expected = new LogisticRegressionAnalysis(inp, outputs);
 
                 expected.Compute();
 
@@ -260,17 +309,8 @@ public void FromSummaryTest1()
             }
 
 
-            {
-                double[][] qtr = data.GetColumn(0).ToDouble().ToArray();
-
-                int[] positive = data.GetColumn(1);
-                int[] negative = data.GetColumn(2);
-
-                actual = LogisticRegressionAnalysis.FromSummary(qtr, positive, negative);
-
-                actual.Compute();
-            }
 
+            var actual = regression;
             Assert.AreEqual(expected.Coefficients[0].Value, actual.Coefficients[0].Value, 1e-8);
             Assert.AreEqual(expected.Coefficients[1].Value, actual.Coefficients[1].Value, 1e-8);