private void UpdateStandardErrors(double[] coefficients, Matrix informationMatrix, int n, int k)
{
for (int index = 0; index < coefficients.Length; index++)
{
var coefficient = coefficients[index];
var standardError = System.Math.Sqrt(informationMatrix.GetValue(index, index).GetNumber());
var tStat = TStatistic(coefficient, standardError);
var pValue = PValue(tStat, n, k);
var coefficientResult = new CoefficientResult
{
Value = coefficient,
StandardError = standardError,
TStatistic = tStat,
PValue = pValue
};
if (index == 0)
{
_linearRegressionResult.Intercept = coefficientResult;
}
else
{
_linearRegressionResult.SlopeLst.Add(coefficientResult);
}
}
}
public MultipleLinearRegression Regress(Matrix matrix, Vector output, bool addOne = true)
{
if (addOne)
{
matrix = matrix.AddColumn(1);
}
output.SetVertical();
if (_alg == MatrixDecompositionAlgs.NAIVE)
{
_singularValueDecomposition.DecomposeNaive(matrix);
}
else
{
_singularValueDecomposition.DecomposeGolubReinsch(matrix);
}
var coefficients = _singularValueDecomposition.Inverse()
.Multiply((Matrix)output)
.Evaluate()
.GetColumnVector(0);
var result = new LinearRegressionResult();
for (int i = 0; i < coefficients.Length; i++)
{
var coefficient = coefficients[i].GetNumber();
var coefficientResult = new CoefficientResult
{
Value = coefficient
};
if (i == 0)
{
result.Intercept = coefficientResult;
}
else
{
result.SlopeLst.Add(coefficientResult);
}
}
_linearRegressionResult = result;
UpdateInformation(matrix, output, addOne);
return(this);
}