private void CalculateParametersAndStatistic(bool calculateStatistics)
{
Matrix XprimeXinv = X.Transpose().Multiply(X).PseudoInverse();
Matrix betaValues = XprimeXinv.Multiply(X.Transpose()).Multiply(Y);
int t = Y.NRows;
int n = betaValues.NRows;
if (calculateStatistics == false)
{
Betas = new ParameterEstimate[n];
for (int i = 0; i < n; i++)
{
Betas[i] = new ParameterEstimate(betaValues[i, 0]);
}
return;
}
Matrix expectedValueY = X.Multiply(betaValues);
Matrix Z = expectedValueY.DeviationsFromMean();
ESS = Z.Transpose().Multiply(Z)[0, 0];
Residuals = expectedValueY.Subtract(Y);
RSS = Residuals.Transpose().Multiply(Residuals)[0, 0];
TrackingError = Math.Sqrt(RSS / (t - 1));
rSquared = 1 - RSS / TSS;
AdjustedRSquared = 1 - (1 - rSquared) * (t - 1) / (t - n);
double varianceOfErrorTerm = Residuals.Transpose().Multiply(Residuals)[0, 0] / (t - n);
Betas = new ParameterEstimate[n];
for (int i = 0; i < n; i++)
{
double sigma = Math.Sqrt(varianceOfErrorTerm * XprimeXinv[i, i]);
double tStat = betaValues[i, 0] == 0 ? 0 : Math.Abs(betaValues[i, 0] - 0) / sigma;
double pValue = 1.0 - Distributions.StudentsTCumulativeDensityFunction(tStat, t - n);
Betas[i] = new ParameterEstimate(betaValues[i, 0], sigma, tStat, pValue);
}
fStatistic = (rSquared / (n - 1)) / ((1 - rSquared) / (t - n));
if (n > 1)
{
fValue = 1.0 - Distributions.FCumulativeDensityFunction(fStatistic, n - 1, t - n);
}
}
