public void CompareIntegerWeights()
{
Random random = new Random((int)DateTime.Now.Ticks);
List <double> xRepeatedValues = new List <double>();
List <double> yRepeatedValues = new List <double>();
List <int> weights = new List <int>();
var sampleTuples = new List <Tuple <double[], double> >();
for (int i = 0; i < 10; i++)
{
int weight = random.Next(1, 10);
weights.Add(weight);
double x = random.NextDouble();
xRepeatedValues.AddRange(Enumerable.Repeat(x, weight));
double y = random.NextDouble();
yRepeatedValues.AddRange(Enumerable.Repeat(y, weight));
sampleTuples.Add(new Tuple <double[], double>(new[] { x }, y));
}
const double epsilon = 1E-12;
var repeatedIntercept = Statistics.Intercept(new Statistics(yRepeatedValues),
new Statistics(xRepeatedValues));
var repeatedSlope = Statistics.Slope(new Statistics(yRepeatedValues), new Statistics(xRepeatedValues));
var repeatedSlopeWithoutIntercept = SlopeWithoutIntercept(yRepeatedValues, xRepeatedValues);
var weightedRegression = WeightedRegression.Weighted(sampleTuples, weights.Select(w => (double)w).ToArray(), true);
Assert.AreEqual(repeatedIntercept, weightedRegression[0], epsilon);
Assert.AreEqual(repeatedSlope, weightedRegression[1], epsilon);
var weightedRegressionWithoutIntercept = WeightedRegression.Weighted(sampleTuples,
weights.Select(w => (double)w).ToArray());
Assert.AreEqual(repeatedSlopeWithoutIntercept, weightedRegressionWithoutIntercept[0], epsilon);
}
public static CalibrationCurve LinearFitThroughZero(IList <WeightedPoint> points)
{
// ReSharper disable RedundantArgumentDefaultValue
var values = WeightedRegression.Weighted(points.Select(p => new Tuple <double[], double>(new[] { p.X }, p.Y)),
points.Select(p => p.Weight).ToArray(), false);
// ReSharper restore RedundantArgumentDefaultValue
return(new CalibrationCurve().ChangePointCount(points.Count).ChangeSlope(values[0]));
}
protected static CalibrationCurve LinearFit(IList <WeightedPoint> points)
{
CalibrationCurve calibrationCurve = new CalibrationCurve().ChangePointCount(points.Count);
try
{
double[] values = WeightedRegression.Weighted(points.Select(p => new Tuple <double[], double>(new[] { p.X }, p.Y)),
points.Select(p => p.Weight).ToArray(), true);
calibrationCurve = calibrationCurve.ChangeSlope(values[1]).ChangeIntercept(values[0]);
}
catch (Exception ex)
{
calibrationCurve = calibrationCurve.ChangeErrorMessage(ex.Message);
}
return(calibrationCurve);
}
/// <summary>
/// Weighted Least-Squares fitting the points (X,y) = ((x0,x1,..,xk),y) and weights w to a linear surface y : X -> p0*x0 + p1*x1 + ... + pk*xk,
/// returning its best fitting parameters as [p0, p1, p2, ..., pk] array.
/// </summary>
public static double[] MultiDimWeighted(double[][] x, double[] y, double[] w)
{
return(WeightedRegression.Weighted(x, y, w));
}
/// <summary>
/// Weighted Least-Squares fitting the points (x,y) and weights w to a k-order polynomial y : x -> p0 + p1*x + p2*x^2 + ... + pk*x^k,
/// returning its best fitting parameters as [p0, p1, p2, ..., pk] array, compatible with Evaluate.Polynomial.
/// A polynomial with order/degree k has (k+1) coefficients and thus requires at least (k+1) samples.
/// </summary>
public static double[] PolynomialWeighted(double[] x, double[] y, double[] w, int order)
{
var design = Matrix <double> .Build.Dense(x.Length, order + 1, (i, j) => Math.Pow(x[i], j));
return(WeightedRegression.Weighted(design, Vector <double> .Build.Dense(y), Matrix <double> .Build.Diagonal(w)).ToArray());
}
