void CalculateValues()
{
#region Values Assigning
Xi_squared = new double[lines - 1];
XiYi = new double[lines - 1];
Xi_m_Xavg_x_Yi_m_YAvg = new double[lines - 1];
Xi_m_X_Avg_Squared = new double[lines - 1];
Yi_m_Y_Avg_Squared = new double[lines - 1];
Y_lineal = new double[lines - 1];
Yi_m_Lineal_Yi_Squared = new double[lines - 1];
Lineal_Yi_m_Y_Avg_Squared = new double[lines - 1];
#endregion
for (int i = 0; i < lines - 1; i++)
{
Xi_squared[i] = Pow(Xi[i], 2);
XiYi[i] = Xi[i] * Yi[i];
SUM_X += Xi[i];
SUM_Y += Yi[i];
SUM_Xi_squared += Xi_squared[i];
SUM_XiYi += XiYi[i];
}
Avg_X = SUM_X / (lines - 1);
Avg_Y = SUM_Y / (lines - 1);
B1 = ((lines - 1) * SUM_XiYi - SUM_X * SUM_Y) / ((lines - 1) * SUM_Xi_squared - Pow(SUM_X, 2));
B0 = SUM_Y / (lines - 1) - B1 * SUM_X / (lines - 1);
for (int i = 0; i < lines - 1; i++)
{
Xi_m_Xavg_x_Yi_m_YAvg[i] = (Xi[i] - Avg_X) * (Yi[i] - Avg_Y);
Xi_m_X_Avg_Squared[i] = Pow(Xi[i] - Avg_X, 2);
Yi_m_Y_Avg_Squared[i] = Pow(Yi[i] - Avg_Y, 2);
Y_lineal[i] = B0 + B1 * Xi[i];
Yi_m_Lineal_Yi_Squared[i] = Pow(Yi[i] - Y_lineal[i], 2);
Lineal_Yi_m_Y_Avg_Squared[i] = Pow(Y_lineal[i] - Avg_Y, 2);
SUM_Xi_m_Xavg_x_Yi_m_YAvg += Xi_m_Xavg_x_Yi_m_YAvg[i];
SUM_Xi_m_X_Avg_Squared += Xi_m_X_Avg_Squared[i];
SUM_Yi_m_Y_Avg_Squared += Yi_m_Y_Avg_Squared[i];
SUM_Y_lineal += Y_lineal[i];
SUM_Yi_m_Lineal_Yi_Squared += Yi_m_Lineal_Yi_Squared[i];
SUM_Lineal_Yi_m_Y_Avg_Squared += Lineal_Yi_m_Y_Avg_Squared[i];
}
Determ_Coef = SUM_Lineal_Yi_m_Y_Avg_Squared / SUM_Yi_m_Y_Avg_Squared;
Indeterm_Coef = 1 - Determ_Coef;
Corelation_P = Sqrt(Determ_Coef);
Corelation_B = SUM_Xi_m_Xavg_x_Yi_m_YAvg / Sqrt(SUM_Xi_m_X_Avg_Squared * SUM_Yi_m_Y_Avg_Squared);
Standard_dev = Sqrt(SUM_Yi_m_Lineal_Yi_Squared / (lines - 3));
Dispersion_1 = SUM_Lineal_Yi_m_Y_Avg_Squared;
Dispersion_2 = SUM_Yi_m_Lineal_Yi_Squared / (lines - 3);
F_em = Dispersion_1 / Dispersion_2;
F_t = new double[2];
F_t[0] = FisherSnedecor.InvCDF(1, lines - 3, 0.95);
F_t[1] = (lines == 4) ? 4052 : FisherSnedecor.InvCDF(1, lines - 3, 0.99);
SendToData();
}
public void ValidateInverseCumulativeDistribution(double d1, double d2, double x)
{
var n = new FisherSnedecor(d1, d2);
double p = SpecialFunctions.BetaRegularized(d1 / 2.0, d2 / 2.0, d1 * x / (d2 + (x * d1)));
Assert.That(n.InverseCumulativeDistribution(p), Is.EqualTo(x).Within(1e-8));
Assert.That(FisherSnedecor.InvCDF(d1, d2, p), Is.EqualTo(x).Within(1e-8));
}
public static double FInv(double probability, int degreesFreedom1, int degreesFreedom2)
{
return(FisherSnedecor.InvCDF(degreesFreedom1, degreesFreedom2, 1d - probability));
}
public (double residualDispersion, double explainedDispersion, double fEmpirical, double fTheoretical, bool isModelAdecuate) CheckAdequacyOfModel()
{
//If the adjustedValues are correct continue with the analysis
if (areAdjustedYValuesAlright == true)
{
double alpha = 0.05D;
double numberOfEquationParameters = 2.0D;
double explainedDeviation = this.GetExplainedDeviation();
double residualDeviation = this.GetResidualDeviation();
double explainedDisperssion = explainedDeviation / (numberOfEquationParameters - 1.0D);
if (Double.IsNaN(explainedDisperssion) || Double.IsInfinity(explainedDisperssion))
{
explainedDisperssion = 1;
}
double residualDisperssion = residualDeviation / (this.YValues.Count - numberOfEquationParameters);
if (Double.IsNaN(residualDisperssion) || Double.IsInfinity(residualDisperssion))
{
residualDisperssion = 1;
}
double FEmpirical = (explainedDisperssion >= residualDisperssion) ? (explainedDisperssion / residualDisperssion) : (residualDisperssion / explainedDisperssion);
if (Double.IsNaN(FEmpirical) || Double.IsInfinity(FEmpirical))
{
FEmpirical = 1;
}
double firstDegreeOfFreedom = 0.0;
double secondDegreeOfFreedom = 0.0;
if (explainedDisperssion > residualDisperssion)
{
firstDegreeOfFreedom = (numberOfEquationParameters - 1);
secondDegreeOfFreedom = (this.Count - numberOfEquationParameters);
}
else if (residualDisperssion > explainedDisperssion)
{
firstDegreeOfFreedom = (this.Count - numberOfEquationParameters);
secondDegreeOfFreedom = (numberOfEquationParameters - 1);
}
else
{
firstDegreeOfFreedom = (numberOfEquationParameters - 1);
secondDegreeOfFreedom = (this.Count - numberOfEquationParameters);
}
if (firstDegreeOfFreedom <= 0)
{
firstDegreeOfFreedom = 1;
}
if (secondDegreeOfFreedom <= 0)
{
secondDegreeOfFreedom = 1;
}
FisherSnedecor fDistibution = new FisherSnedecor(1, 1);
double FTheoretical = FisherSnedecor.InvCDF(firstDegreeOfFreedom, secondDegreeOfFreedom, (1.00 - alpha));
bool isModelAdequate = false;
if (FEmpirical <= FTheoretical)
{
isModelAdequate = false;
}
else if (FEmpirical > FTheoretical)
{
isModelAdequate = true;
}
return(residualDisperssion, explainedDisperssion, FEmpirical, FTheoretical, isModelAdequate);
}
else
{
//Else return default values.
return(0.0, 0.0, 0.0, 0.0, false);
}
}
/// <summary>
/// Inverzna Fisherova (F) distribucija (razdioba)
/// </summary>
/// <param name="alfa">Nivo signifikantnosti (znacajnosti) -> 1 > alfa > 0 </param>
/// <param name="f1">Stupanj slobode</param>
/// <param name="f2">Stupanj slobode (ako uvrstite veliki broj dvostruka Fisherova prelazi u Chi^2)</param>
/// <returns>double</returns>
public static double Fisher(double alfa, int f1, int f2)
{
return(FisherSnedecor.InvCDF(f1, f2, 1 - alfa));
}
