640 lines
22 KiB
MQL5
640 lines
22 KiB
MQL5
|
//+------------------------------------------------------------------+
|
||
|
|
//| EconometricsA.mqh |
|
||
|
|
//| Copyright 2000-2026, MetaQuotes Ltd. |
|
||
|
|
//| www.mql5.com |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
#include <Math\Stat\T.mqh>
|
||
|
|
#include <Math\Stat\ChiSquare.mqh>
|
||
|
|
#include <Math\Stat\F.mqh>
|
||
|
|
#include <Math\Stat\Math.mqh>
|
||
|
|
#include <Graphics\Graphic.mqh>
|
||
|
|
//---
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Computation of parameters and residuals |
|
||
|
|
//| for regression on a constant |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void regression0(double& y[], double& a, double& c, double& e[])
|
||
|
|
{
|
||
|
|
//--- Initialization and input data validation
|
||
|
|
a = c = 0.0;
|
||
|
|
ArrayResize(e, 0);
|
||
|
|
int ny = ArraySize(y);
|
||
|
|
if(ny < 1)
|
||
|
|
{
|
||
|
|
Print("no data for regression0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
ArrayResize(e, ny);
|
||
|
|
//--- Parameters computation
|
||
|
|
a = MathMean(y);
|
||
|
|
c = MathStandardDeviation(y);
|
||
|
|
//--- Residuals computation
|
||
|
|
for(int i = 0; i < ny; ++i)
|
||
|
|
e[i] = y[i] - a;
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Calculation of confidence interval for parameter A |
|
||
|
|
//| in regression on a constant |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void Aconf0(double a, double c, int n, double conf_level)
|
||
|
|
{
|
||
|
|
//--- Initialization and input data validation
|
||
|
|
if(c <= 0.0)
|
||
|
|
{
|
||
|
|
Print("c must be positive in Aconf0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(n < 2)
|
||
|
|
{
|
||
|
|
Print("n must be greater 1 in Aconf0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(conf_level <= 0.0 || conf_level >= 1.0)
|
||
|
|
{
|
||
|
|
Print("conf_level must be between 0 and 1 in Aconf0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
int err = 0;
|
||
|
|
//--- t-statistic calculation
|
||
|
|
//--- double t = MathQuantileT((1.0 - conf_level) / 2.0, n - 1, err); // faulty library function
|
||
|
|
double t = MathQuantileT_TMP((1.0 - conf_level) / 2.0, n - 1, err); // custom implementation
|
||
|
|
if(err != 0)
|
||
|
|
{
|
||
|
|
Print("in MathQuantileT() error ", err);
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
t *= c / MathSqrt(n);
|
||
|
|
//--- Print result
|
||
|
|
PrintFormat("A between %.4f and %.4f with confidence level %.3f", a + t, a - t, conf_level);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Test of hypothesis H₀: A = 0 vs H₁: A > 0 |
|
||
|
|
//| for regression on a constant |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void Aeq0_test0(double a, double c, int n)
|
||
|
|
{
|
||
|
|
Print("test H0: A = 0 vs H1: A > 0 (for regession on constant) result:");
|
||
|
|
//--- Input data validation
|
||
|
|
if(c <= 0.0)
|
||
|
|
{
|
||
|
|
Print("c must be positive in Aeq0_test0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(n < 2)
|
||
|
|
{
|
||
|
|
Print("n must be greater 1 in Aeq0_test0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- t-statistic calculation
|
||
|
|
double t = a * MathSqrt(n) / c;
|
||
|
|
int err = 0;
|
||
|
|
//--- p-value calculation
|
||
|
|
double p_value = MathCumulativeDistributionT(t, n - 1, false, false, err);
|
||
|
|
if(err != 0)
|
||
|
|
{
|
||
|
|
Print("in MathCumulativeDistributionT() error ", err);
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Print result if success
|
||
|
|
PrintFormat("t = %.3f, p-value = %.3f", t, p_value);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Calculation of prediction interval for Y |
|
||
|
|
//| in regression on a constant |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void prognose_interval0(double a, double c, int n, double conf_level)
|
||
|
|
{
|
||
|
|
//--- Input data validation
|
||
|
|
if(c <= 0.0)
|
||
|
|
{
|
||
|
|
Print("c must be positive in prognose_interval0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(n < 2)
|
||
|
|
{
|
||
|
|
Print("n must be greater 1 in prognose_interval0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(conf_level <= 0.0 || conf_level >= 1.0)
|
||
|
|
{
|
||
|
|
Print("conf_level must be between 0 and 1 in prognose_interval0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
int err = 0;
|
||
|
|
//--- t-statistic calculation
|
||
|
|
//--- double t = MathQuantileT((1.0 - conf_level) / 2.0, n - 1, err); // faulty library function
|
||
|
|
double t = MathQuantileT_TMP((1.0 - conf_level) / 2.0, n - 1, err); // custom implementation
|
||
|
|
if(err != 0)
|
||
|
|
{
|
||
|
|
Print("in MathQuantileT() error ", err);
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
t *= c * MathSqrt(1.0 + 1.0 / n);
|
||
|
|
//--- Print result if success
|
||
|
|
PrintFormat("Y between %.4f and %.4f with confidence level %.3f", a + t, a - t, conf_level);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Computation of parameters and residuals for |
|
||
|
|
//| simple linear regression |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void regression1(double& y[], double& x[], double& a, double& b, double& c, double& e[])
|
||
|
|
{
|
||
|
|
//--- Initialization and input data validation
|
||
|
|
a = b = c = 0.0;
|
||
|
|
ArrayResize(e, 0);
|
||
|
|
int ny = ArraySize(y);
|
||
|
|
if(ny < 3)
|
||
|
|
{
|
||
|
|
Print("no data for regression1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(ny != ArraySize(x))
|
||
|
|
{
|
||
|
|
Print("arrays y[] and x[] have different lengths in regression1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Check x[] is not constant
|
||
|
|
double Sx = MathStandardDeviation(x);
|
||
|
|
if(Sx <= DBL_MIN)
|
||
|
|
{
|
||
|
|
Print("x[]==const, use regression0()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Parameters A and B estimation
|
||
|
|
if(!MathCorrelationPearson(y, x, b))
|
||
|
|
{
|
||
|
|
Print("MathCorrelationPearson() error in regression1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
ArrayResize(e, ny);
|
||
|
|
b *= MathStandardDeviation(y) / Sx;
|
||
|
|
a = MathMean(y) - b * MathMean(x);
|
||
|
|
//--- Residuals computation
|
||
|
|
for(int i = 0; i < ny; ++i)
|
||
|
|
e[i] = y[i] - a - b * x[i];
|
||
|
|
//--- Parameter C estimation
|
||
|
|
c = MathStandardDeviation(e) * MathSqrt(1.0 + 1.0 / (ny - 2.0));
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Test of hypothesis H₀: B = 0 vs H₁: B ≠ 0 for |
|
||
|
|
//| simple linear regression |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void Beq0_test1(double b, double c, double var_x, int n)
|
||
|
|
{
|
||
|
|
Print("test H0: B = 0 vs H1: B != 0 (for pair regession) result:");
|
||
|
|
//--- Input data validation
|
||
|
|
if(c <= 0.0)
|
||
|
|
{
|
||
|
|
Print("c must be positive in Beq0_test1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(var_x <= 0.0)
|
||
|
|
{
|
||
|
|
Print("var_x must be positive in Beq0_test1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(n < 3)
|
||
|
|
{
|
||
|
|
Print("n must be 3 or greater in Beq0_test1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- t-statistic calculation
|
||
|
|
double t = MathAbs(b * MathSqrt(var_x * (n - 1)) / c);
|
||
|
|
int err = 0;
|
||
|
|
//--- p-value computation
|
||
|
|
double p_value = 2.0 * MathCumulativeDistributionT(t, n - 2, false, false, err);
|
||
|
|
if(err != 0)
|
||
|
|
{
|
||
|
|
Print("in MathCumulativeDistributionT() error ", err);
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Print result if success
|
||
|
|
PrintFormat("t = %.3f, p-value = %.3f", t, p_value);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Calculation of prediction interval for Y in |
|
||
|
|
//| simple linear regression |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void prognose_interval1(double x_new, double a, double b, double c, double var_x, double mean_x, int n, double conf_level)
|
||
|
|
{
|
||
|
|
//--- Input data validation
|
||
|
|
if(c <= 0.0)
|
||
|
|
{
|
||
|
|
Print("c must be positive in prognose_interval1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(n < 2)
|
||
|
|
{
|
||
|
|
Print("n must be greater 2 in prognose_interval1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(conf_level <= 0.0 || conf_level >= 1.0)
|
||
|
|
{
|
||
|
|
Print("conf_level must be between 0 and 1 in prognose_interval1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
int err = 0;
|
||
|
|
//--- t-statistic calculation
|
||
|
|
//--- double t = MathQuantileT((1.0 - conf_level) / 2.0, n - 1, err); // faulty library function
|
||
|
|
double t = MathQuantileT_TMP((1.0 - conf_level) / 2.0, n - 1, err); // custom implementation
|
||
|
|
if(err != 0)
|
||
|
|
{
|
||
|
|
Print("in MathQuantileT() error ", err);
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
t *= c * MathSqrt(1.0 + 1.0 / n + MathPow((x_new - mean_x), 2) * (n - 1) / var_x);
|
||
|
|
//--- Print result if success
|
||
|
|
PrintFormat("Y between %.4f and %.4f with confidence level %.3f", a + b * x_new + t, a + b * x_new - t, conf_level);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Residuals plot vs. price bar index |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void t_residuals_plot(double& residuals[])
|
||
|
|
{
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
//--- Input data validation
|
||
|
|
if(n_residuals < 1)
|
||
|
|
{
|
||
|
|
Print("no data for t_residuals_plot()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Array T[] as x on plot
|
||
|
|
double T[];
|
||
|
|
if(!MathSequenceByCount(1, n_residuals, n_residuals, T))
|
||
|
|
{
|
||
|
|
Print("MathSequenceByCount() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Plot construction
|
||
|
|
ChartSetInteger(0, CHART_SHOW, false);
|
||
|
|
CGraphic graphic;
|
||
|
|
graphic.Create(0, "G", 0, 0, 0, 750, 300);
|
||
|
|
graphic.CurveAdd(T, residuals, CURVE_LINES, "Residuals");
|
||
|
|
graphic.CurvePlotAll();
|
||
|
|
graphic.Update();
|
||
|
|
MessageBox("Press OK to close graph", "Close graph");
|
||
|
|
ChartSetInteger(0, CHART_SHOW, true);
|
||
|
|
graphic.Destroy();
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| EPDF of residuals vs. normal density with residual SD |
|
||
|
|
//| nofx - number of points on plot |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void epdf_vs_normalpdf(double& residuals[], int nofx = 30)
|
||
|
|
{
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
//--- Input data validation
|
||
|
|
if(n_residuals < 1)
|
||
|
|
{
|
||
|
|
Print("no data for epdf_vs_normalpdf()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Array for x
|
||
|
|
double x[];
|
||
|
|
if(!MathSequenceByCount(MathMin(residuals), MathMax(residuals), nofx, x))
|
||
|
|
{
|
||
|
|
Print("sequence error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Array for EPDF
|
||
|
|
double epdf[];
|
||
|
|
if(!MathProbabilityDensityEmpirical(residuals, nofx, x, epdf))
|
||
|
|
{
|
||
|
|
Print("MathProbabilityDensityEmpirical() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Array for theoretical PDF
|
||
|
|
double normal_pdf[];
|
||
|
|
if(!MathProbabilityDensityNormal(x, 0.0, MathStandardDeviation(residuals), normal_pdf))
|
||
|
|
{
|
||
|
|
Print("MathProbabilityDensityNormal() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Plot construction
|
||
|
|
ChartSetInteger(0, CHART_SHOW, false);
|
||
|
|
CGraphic graphic;
|
||
|
|
graphic.Create(0, "G", 0, 0, 0, 750, 400);
|
||
|
|
graphic.CurveAdd(x, epdf, CURVE_LINES, "Residuals");
|
||
|
|
graphic.CurveAdd(x, normal_pdf, CURVE_LINES, "Normal");
|
||
|
|
graphic.CurvePlotAll();
|
||
|
|
graphic.Update();
|
||
|
|
MessageBox("Press OK to close graph", "Close graph");
|
||
|
|
ChartSetInteger(0, CHART_SHOW, true);
|
||
|
|
graphic.Destroy();
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| QQ-plot of residuals vs. normal distribution with residual SD |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void qq_plot(double& residuals[])
|
||
|
|
{
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
//--- Input data validation
|
||
|
|
if(n_residuals < 1)
|
||
|
|
{
|
||
|
|
Print("no data for qq_plot()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Copy input array and sort this copy
|
||
|
|
double residuals_copy[];
|
||
|
|
ArrayResize(residuals_copy, n_residuals);
|
||
|
|
ArrayCopy(residuals_copy, residuals);
|
||
|
|
ArraySort(residuals_copy);
|
||
|
|
//--- Probability sequence
|
||
|
|
double pseq[];
|
||
|
|
if(!MathSequenceByCount(0.5 / n_residuals, 1.0 - 0.5 / n_residuals, n_residuals, pseq))
|
||
|
|
{
|
||
|
|
Print("MathSequenceByCount() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Theoretical quantiles
|
||
|
|
double normal_qs[];
|
||
|
|
if(!MathQuantileNormal(pseq, 0.0, MathStandardDeviation(residuals_copy), normal_qs))
|
||
|
|
{
|
||
|
|
Print("MathQuantileNormal() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Plot construction
|
||
|
|
ChartSetInteger(0, CHART_SHOW, false);
|
||
|
|
CGraphic graphic;
|
||
|
|
graphic.Create(0, "G", 0, 0, 0, 750, 550);
|
||
|
|
graphic.CurveAdd(normal_qs, residuals_copy, CURVE_LINES, "QQ-plot");
|
||
|
|
graphic.CurveAdd(residuals_copy, residuals_copy, CURVE_LINES, "Y=X");
|
||
|
|
graphic.CurvePlotAll();
|
||
|
|
graphic.Update();
|
||
|
|
MessageBox("Press OK to close graph", "Close graph");
|
||
|
|
ChartSetInteger(0, CHART_SHOW, true);
|
||
|
|
graphic.Destroy();
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Correlogram of residuals |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void correlogram(double& residuals[])
|
||
|
|
{
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
//--- Input data validation
|
||
|
|
if(n_residuals < 2)
|
||
|
|
{
|
||
|
|
Print("no data for correlogram()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Number of lags
|
||
|
|
int ncg = (int)MathSqrt(n_residuals);
|
||
|
|
//--- Significance levels
|
||
|
|
double bound = 1.96 / MathSqrt(n_residuals);
|
||
|
|
double cg0 = 0.0, xb[2] = {0, ncg + 1}, upb[2] = {bound, bound}, lowb[2] = {-bound, -bound};
|
||
|
|
//--- Lags array creation
|
||
|
|
double lags[];
|
||
|
|
if(!MathSequenceByCount(1, ncg, ncg, lags))
|
||
|
|
{
|
||
|
|
Print("MathSequenceByCount() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- ACF array calculation
|
||
|
|
double cg[];
|
||
|
|
ArrayResize(cg, ncg);
|
||
|
|
ArrayFill(cg, 0, ncg, 0.0);
|
||
|
|
for(int i = 0; i < n_residuals; ++i)
|
||
|
|
cg0 += residuals[i] * residuals[i];
|
||
|
|
for(int k = 1; k <= ncg; ++k)
|
||
|
|
{
|
||
|
|
for(int i = 0; i < n_residuals - k; ++i)
|
||
|
|
cg[k - 1] += residuals[i] * residuals[i + k];
|
||
|
|
cg[k - 1] /= cg0;
|
||
|
|
}
|
||
|
|
//--- Plot construction
|
||
|
|
ChartSetInteger(0, CHART_SHOW, false);
|
||
|
|
CGraphic graphic;
|
||
|
|
graphic.Create(0, "G", 0, 0, 0, 750, 300);
|
||
|
|
graphic.CurveAdd(lags, cg, CURVE_HISTOGRAM, "Correlogram");
|
||
|
|
graphic.CurveAdd(xb, upb, CURVE_LINES, "Upper bound");
|
||
|
|
graphic.CurveAdd(xb, lowb, CURVE_LINES, "Lower bound");
|
||
|
|
graphic.CurvePlotAll();
|
||
|
|
graphic.Update();
|
||
|
|
MessageBox("Press OK to close graph", "Close graph");
|
||
|
|
ChartSetInteger(0, CHART_SHOW, true);
|
||
|
|
graphic.Destroy();
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Scatter plot of (x, y) points with line y = a * x + b |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void scatter_plot(double& x[], double& y[], bool add_line = false, double a = 0.0, double b = 0.0)
|
||
|
|
{
|
||
|
|
int ny = ArraySize(y);
|
||
|
|
//--- Input data validation
|
||
|
|
if(ny < 1)
|
||
|
|
{
|
||
|
|
Print("no data for regression1()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(ny != ArraySize(x))
|
||
|
|
{
|
||
|
|
Print("arrays y[] and x[] have different lengths in scatter_plot()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Regression line points calculation
|
||
|
|
double xreg[2] = {MathMin(x), MathMax(x)};
|
||
|
|
double yreg[2] = {a + b*xreg[0], a + b*xreg[1]};
|
||
|
|
//--- Plot construction
|
||
|
|
ChartSetInteger(0, CHART_SHOW, false);
|
||
|
|
CGraphic graphic;
|
||
|
|
graphic.Create(0, "G", 0, 0, 0, 750, 550);
|
||
|
|
graphic.CurveAdd(x, y, CURVE_POINTS, "scatter plot");
|
||
|
|
if(add_line)
|
||
|
|
graphic.CurveAdd(xreg, yreg, CURVE_LINES, "y = a * x + b");
|
||
|
|
graphic.CurvePlotAll();
|
||
|
|
graphic.Update();
|
||
|
|
MessageBox("Press OK to close graph", "Close graph");
|
||
|
|
ChartSetInteger(0, CHART_SHOW, true);
|
||
|
|
graphic.Destroy();
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Ljung-Box test for autocorrelation in residuals |
|
||
|
|
//| m - lags number, pq - sum of p and q in ARMA(p,q) model |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void Ljung_Box_test(double& residuals[], int m = 10, int pq = 0)
|
||
|
|
{
|
||
|
|
Print("Ljung-Box test result:");
|
||
|
|
//--- Input data validation
|
||
|
|
if(pq < 0)
|
||
|
|
{
|
||
|
|
Print("pq must be nonnegative");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
if(m <= pq)
|
||
|
|
{
|
||
|
|
Print("m must be greater pq");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
if(n_residuals < m + 1)
|
||
|
|
{
|
||
|
|
Print("not enough data");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Initialization
|
||
|
|
double Q = 0.0, cov0 = 0.0, cov, ro;
|
||
|
|
//--- Test statistic calculation
|
||
|
|
for(int i = 0; i < n_residuals; ++i)
|
||
|
|
cov0 += residuals[i] * residuals[i];
|
||
|
|
for(int k = 1; k <= m; ++k)
|
||
|
|
{
|
||
|
|
cov = 0.0;
|
||
|
|
for(int i = 0; i < n_residuals - k; ++i)
|
||
|
|
cov += residuals[i] * residuals[i + k];
|
||
|
|
ro = cov / cov0;
|
||
|
|
Q += ro * ro / (n_residuals - k);
|
||
|
|
}
|
||
|
|
Q *= n_residuals * (n_residuals + 2);
|
||
|
|
//--- p-value computation
|
||
|
|
int err = 0;
|
||
|
|
double p_value = MathCumulativeDistributionChiSquare(Q, m - pq, false, false, err);
|
||
|
|
if(err != 0)
|
||
|
|
{
|
||
|
|
Print("MathCumulativeDistributionChiSquare() error: ", err);
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Print result
|
||
|
|
PrintFormat("Q = %.3f, p-value = %.3f", Q, p_value);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Jarque-Bera test for normality of distribution |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void Jarque_Bera_test(double& residuals[])
|
||
|
|
{
|
||
|
|
Print("Jarque-Bera test result:");
|
||
|
|
//--- Input data validation
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
if(n_residuals < 1)
|
||
|
|
{
|
||
|
|
Print("not enough data");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Skewness calculation
|
||
|
|
double S = MathSkewness(residuals);
|
||
|
|
//--- Kurtosis calculation
|
||
|
|
double K = MathKurtosis(residuals);
|
||
|
|
//--- Test statistic calculation
|
||
|
|
double JB = n_residuals * (S * S + K * K / 4.0) / 6.0;
|
||
|
|
//--- p-value computation
|
||
|
|
double p_value = MathExp(-JB / 2.0);
|
||
|
|
//--- Print result
|
||
|
|
PrintFormat("JB = %.3f, p-value = %.3f", JB, p_value);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Pettitt test for structural break (abrupt change in mean) |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void Pettitt_test(double& residuals[])
|
||
|
|
{
|
||
|
|
Print("Pettitt test result:");
|
||
|
|
//--- Input data validation
|
||
|
|
int n_residuals = ArraySize(residuals);
|
||
|
|
if(n_residuals < 2)
|
||
|
|
{
|
||
|
|
Print("not enough data");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Initialization
|
||
|
|
double K = 0.0, U, W = 0.0;
|
||
|
|
int t_change = 0;
|
||
|
|
//--- Sample ranking computation
|
||
|
|
double rank[];
|
||
|
|
MathRank(residuals, rank);
|
||
|
|
//--- Test statistic calculation
|
||
|
|
for(int t = 1; t < n_residuals; ++t)
|
||
|
|
{
|
||
|
|
W += rank[t - 1];
|
||
|
|
U = MathAbs(2.0 * W - t * (n_residuals + 1.0));
|
||
|
|
if(U > K)
|
||
|
|
{
|
||
|
|
K = U;
|
||
|
|
t_change = t;
|
||
|
|
}
|
||
|
|
}
|
||
|
|
//--- p-value computation
|
||
|
|
double p_value = MathMin(2.0 * MathExp(-6.0 * K * K / (n_residuals * n_residuals * (n_residuals + 1.0))), 1.0);
|
||
|
|
//--- Print result
|
||
|
|
PrintFormat("K = %.3f, p-value = %.3f, change-point = %d", K, p_value, t_change);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Student's t-distribution quantile |
|
||
|
|
//| (custom implementation replacing faulty library function) |
|
||
|
|
//| JUST A TEMPORARY REPLACEMENT! (Based on library variant) |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
double MathQuantileT_TMP(const double prob, const double nu, int &error_code)
|
||
|
|
{
|
||
|
|
if(!MathIsValidNumber(prob) || !MathIsValidNumber(nu))
|
||
|
|
{
|
||
|
|
error_code = ERR_ARGUMENTS_NAN;
|
||
|
|
return QNaN;
|
||
|
|
}
|
||
|
|
if(nu <= 0.0 || nu != MathRound(nu))
|
||
|
|
{
|
||
|
|
error_code = ERR_ARGUMENTS_INVALID;
|
||
|
|
return QNaN;
|
||
|
|
}
|
||
|
|
if(prob < 0.0 || prob > 1.0)
|
||
|
|
{
|
||
|
|
error_code = ERR_ARGUMENTS_INVALID;
|
||
|
|
return QNaN;
|
||
|
|
}
|
||
|
|
error_code = ERR_RESULT_INFINITE;
|
||
|
|
if(prob == 0.0)
|
||
|
|
return QNEGINF;
|
||
|
|
if(prob == 1.0)
|
||
|
|
return QPOSINF;
|
||
|
|
error_code = ERR_OK;
|
||
|
|
if(prob == 0.5)
|
||
|
|
return 0.0;
|
||
|
|
if(nu == 1.0)
|
||
|
|
return MathTan(M_PI * (prob - 0.5));
|
||
|
|
int err_code = 0;
|
||
|
|
double x_low = -1.0;
|
||
|
|
double x_high = 1.0;
|
||
|
|
while(MathCumulativeDistributionT(x_low, nu, err_code) > prob && x_low > -1e6)
|
||
|
|
x_low *= 2.0;
|
||
|
|
while(MathCumulativeDistributionT(x_high, nu, err_code) < prob && x_high < 1e6)
|
||
|
|
x_high *= 2.0;
|
||
|
|
double x = 0.0;
|
||
|
|
int max_iterations = 100;
|
||
|
|
double precision = 1e-12; // Высокая точность
|
||
|
|
for(int i = 0; i < max_iterations; i++)
|
||
|
|
{
|
||
|
|
x = (x_low + x_high) / 2.0;
|
||
|
|
double cdf = MathCumulativeDistributionT(x, nu, err_code);
|
||
|
|
if(MathAbs(x_high - x_low) < precision)
|
||
|
|
break;
|
||
|
|
if(cdf < prob)
|
||
|
|
x_low = x;
|
||
|
|
else
|
||
|
|
x_high = x;
|
||
|
|
if(i == max_iterations - 1)
|
||
|
|
error_code = ERR_NON_CONVERGENCE;
|
||
|
|
}
|
||
|
|
return x;
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
//| Helper function to export array to text file |
|
||
|
|
//| (for data analysis in other programs) |
|
||
|
|
//+------------------------------------------------------------------+
|
||
|
|
void array2csv(double& a[], string fldr, string fnm)
|
||
|
|
{
|
||
|
|
//--- Initialization and input data validation
|
||
|
|
int ftxt = FileOpen(fldr + "\\" + fnm, FILE_WRITE | FILE_TXT | FILE_ANSI | FILE_COMMON);
|
||
|
|
if(ftxt == INVALID_HANDLE)
|
||
|
|
{
|
||
|
|
Print("FileOpen() error");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
int na = ArraySize(a);
|
||
|
|
if(na < 1)
|
||
|
|
{
|
||
|
|
Print("No data for array2csv()");
|
||
|
|
return;
|
||
|
|
}
|
||
|
|
//--- Writing array to file
|
||
|
|
FileWriteString(ftxt, DoubleToString(a[0]));
|
||
|
|
for(int i = 1; i < na; ++i)
|
||
|
|
FileWriteString(ftxt, "\n" + DoubleToString(a[i]));
|
||
|
|
FileClose(ftxt);
|
||
|
|
}
|
||
|
|
//+------------------------------------------------------------------+
|