示例#1
0
        public static void TestMean(IGaussianDistribution <double> dist)
        {
            int sampleCount = 10000000;

            double sum = 0.0;

            for (int i = 0; i < sampleCount; i++)
            {
                sum += dist.Sample();
            }

            double mean = sum / sampleCount;

            Assert.IsTrue(Math.Abs(mean) < 0.001);
        }
示例#2
0
        public static void TestSimpleStats(IGaussianDistribution <double> dist)
        {
            const int sampleCount = 20_000_000;

            RunningStatistics runningStats = new RunningStatistics();

            for (int i = 0; i < sampleCount; i++)
            {
                runningStats.Push(dist.Sample());
            }

            Assert.IsTrue(Math.Abs(runningStats.Mean) < 0.001);
            Assert.IsTrue(Math.Abs(runningStats.StandardDeviation - 1.0) < 0.0005);
            Assert.IsTrue(Math.Abs(runningStats.Skewness) < 0.01);
            Assert.IsTrue(Math.Abs(runningStats.Kurtosis) < 0.01);
        }
示例#3
0
        public static void TestStandardDeviation(IGaussianDistribution <double> dist)
        {
            int sampleCount = 10000000;

            double sqrSum = 0.0;

            for (int i = 0; i < sampleCount; i++)
            {
                double x = dist.Sample();
                sqrSum += x * x;
            }

            double var    = sqrSum / sampleCount;
            double stdDev = Math.Sqrt(var);

            Assert.IsTrue(Math.Abs(stdDev - 1.0) < 0.001);
        }
示例#4
0
        public static void TestDistribution(IGaussianDistribution <double> dist, double mean, double stdDev)
        {
            // Take a set of samples.
            const int sampleCount = 10_000_000;

            double[] sampleArr = new double[sampleCount];

            for (int i = 0; i < sampleCount; i++)
            {
                sampleArr[i] = dist.Sample();
            }

            // Sort the ample so that we can use SortedArrayStatistics.
            Array.Sort(sampleArr);

            //// Test a range of centile/quantile values.
            double lowerBound = -5;
            double upperBound = 5;

            double tauStep = (upperBound - lowerBound) / 30.0;

            for (double tau = 0; tau <= 1.0; tau += 0.1)
            {
                // Notes.
                // Here we calc the tau'th quartile over a range of values in he interval [0,1],
                // the resulting quantile is the sample value (and CDF x-axis value) at which the
                // CDF crosses tau on the y-axis.
                //
                // We then take that sample x-axis value, pass it through the CDF function for the
                // gaussian to obtain the expected y value at that x, and compare with tau.

                // Determine the x value at which tau (as a proportion) of samples are <= x.
                double sample_x = SortedArrayStatistics.Quantile(sampleArr, tau);

                // Put sample_x into the gaussian CDF function, to obtain a CDF y coord..
                double cdf_y = 0.5 * SpecialFunctions.Erfc((mean - sample_x) / (stdDev * Constants.Sqrt2));

                // Compare the expected and actual CDF y values.
                double y_error = Math.Abs(tau - cdf_y);
                Assert.IsTrue(y_error < 0.0005);
            }
        }