SSJ
3.3.1
Stochastic Simulation in Java
|
Extends the class NortaInitDisc and implements the algorithm NI3. More...
Public Member Functions | |
NI3 (double rX, DiscreteDistributionInt dist1, DiscreteDistributionInt dist2, double tr, double tolerance) | |
Constructor with the target rank correlation rX , the two discrete marginals dist1 and dist2 , the parameter for the truncation tr (see the constructor of class NortaInitDisc ), and the specific parameter \(\epsilon=\) tolerance for the algorithm NI3, as defined above. | |
double | computeCorr () |
Computes and returns the correlation \(\rho_Z\) using algorithm NI3. More... | |
String | toString () |
Public Member Functions inherited from NortaInitDisc | |
NortaInitDisc (double rX, DiscreteDistributionInt dist1, DiscreteDistributionInt dist2, double tr) | |
Constructor with the target rank correlation rX , the two discrete marginals dist1 and dist2 and the parameter for the truncation tr . More... | |
abstract double | computeCorr () |
Computes and returns the correlation \(\rho_Z\). More... | |
void | computeParams () |
Computes the following inputs of each marginal distribution: More... | |
double | integ (double r) |
Computes the function. More... | |
double | deriv (double r) |
Computes the derivative of \(g_r\), given by. More... | |
String | toString () |
Additional Inherited Members | |
Protected Attributes inherited from NortaInitDisc | |
double | rX |
DiscreteDistributionInt | dist1 |
DiscreteDistributionInt | dist2 |
double | tr |
double | mu1 |
Package Attributes inherited from NortaInitDisc | |
double | mu2 |
double | sd1 |
double | sd2 |
int | m2 |
Extends the class NortaInitDisc and implements the algorithm NI3.
It uses the function \(g_r\) and its derivative \(gā_r\), so it calls the methods integ and deriv, given in ( gr_M ) and ( grp_M ) and uses an adapted version of the Newton-Raphson algorithm. At any iteration, if the solution falls outside the search interval, the algorithm uses bisection and halves the interval length to guarantee convergence. The initial solution is taken as \(\rho_0=2 \sin(\pi r_X /6)\), and then at each iteration \(k\), \(f_r(\rho_k)\) and \(f_r^ā(\rho_k)\) are calculated and a solution is computed by the recurrence formula:
\[ \rho_{k+1}=\rho_k-\frac{f_r(\rho_k)}{f_r^ā(\rho_k)}. \]
The algorithm stops at iteration \(k\) if \(|\rho_{k-1}-\rho_k|\leq\epsilon\). The function \(f_r\) is the one given in ( fr ).
double computeCorr | ( | ) |
Computes and returns the correlation \(\rho_Z\) using algorithm NI3.
f, df, dx correspond to f(rho_k), f'(rho_k) and f(rho_k)/f'(rho_k), respectively, in the paper (paragraph "Method NI3"of section 3).