/** Computes the probability <SPAN CLASS="MATH"><I>p</I>(<I>x</I>)</SPAN>. */
public static double prob(double n, double p, int x) {
final int SLIM = 15; // To avoid overflow
final double MAXEXP = (Num.DBL_MAX_EXP - 1) * Num.LN2; // To avoid overflow
final double MINEXP = (Num.DBL_MIN_EXP - 1) * Num.LN2; // To avoid underflow
double y;
if (p < 0.0 || p > 1.0) throw new IllegalArgumentException("p not in [0, 1]");
if (n <= 0.0) throw new IllegalArgumentException("n <= 0.0");
if (x < 0) return 0.0;
if (p >= 1.0) { // In fact, p == 1
if (x == 0) return 1.0;
else return 0.0;
}
if (p <= 0.0) // In fact, p == 0
return 0.0;
y =
Num.lnGamma(n + x)
- (Num.lnFactorial(x) + Num.lnGamma(n))
+ n * Math.log(p)
+ x * Math.log1p(-p);
if (y >= MAXEXP) throw new IllegalArgumentException("term overflow");
return Math.exp(y);
}
/** Computes the density function @f$f(x)@f$. */
public static double density(double alpha, double lambda, double x) {
if (lambda <= 0) throw new IllegalArgumentException("lambda <= 0");
double z = lambda * (x - alpha);
if (z >= -100.0) {
double v = Math.exp(-z);
return lambda * v / ((1.0 + v) * (1.0 + v));
}
return lambda * Math.exp(z);
}
/**
* Estimates the parameters @f$(\alpha, \lambda)@f$ of the logistic distribution using the maximum
* likelihood method, from the @f$n@f$ observations @f$x[i]@f$, @f$i = 0, 1,…, n-1@f$. The
* estimates are returned in a two-element array, in regular order: [@f$\alpha@f$,
*
* @f$\lambda@f$]. The maximum likelihood estimators are the values
* @f$(\hat{\alpha}, \hat{\lambda})@f$ that satisfy the equations:
* @f{align*}{ \sum_{i=1}^n \frac{1}{1 + e^{\hat{\lambda} (x_i - \hat{\alpha})}} & = \frac{n}{2}
* \\ \sum_{i=1}^n \hat{\lambda} (x_i - \hat{\alpha}) \frac{1 - e^{\hat{\lambda} (x_i -
* \hat{\alpha})}}{1 + e^{\hat{\lambda} (x_i - \hat{\alpha})}} & = n.
* @f} See @cite mEVA00a  (page 128).
* @param x the list of observations used to evaluate parameters
* @param n the number of observations used to evaluate parameters
* @return returns the parameter [@f$\hat{\alpha}@f$,
* @f$\hat{\lambda}@f$]
*/
public static double[] getMLE(double[] x, int n) {
if (n <= 0) throw new IllegalArgumentException("n <= 0");
double sum = 0.0;
for (int i = 0; i < n; i++) sum += x[i];
double[] param = new double[3];
param[1] = sum / (double) n;
sum = 0.0;
for (int i = 0; i < n; i++) sum += ((x[i] - param[1]) * (x[i] - param[1]));
param[2] = Math.sqrt(Math.PI * Math.PI * n / (3.0 * sum));
double[] fvec = new double[3];
double[][] fjac = new double[3][3];
int[] iflag = new int[2];
int[] info = new int[2];
int[] ipvt = new int[3];
Optim system = new Optim(x, n);
Minpack_f77.lmder1_f77(system, 2, 2, param, fvec, fjac, 1e-5, info, ipvt);
double parameters[] = new double[2];
parameters[0] = param[1];
parameters[1] = param[2];
return parameters;
}
public double evaluate(double gam) {
if (gam <= 0) return 1.0e100;
double sum = 0.0;
for (int j = 0; j < m; j++) sum += Num.digamma(gam + x[j]);
return sum / m + Math.log(p) - Num.digamma(gam);
}
/** Computes the inverse distribution function @f$F^{-1}(u)@f$. */
public static double inverseF(double alpha, double lambda, double u) {
if (lambda <= 0) throw new IllegalArgumentException("lambda <= 0");
if (u < 0.0 || u > 1.0) throw new IllegalArgumentException("u not in [0, 1]");
if (u >= 1.0) return Double.POSITIVE_INFINITY;
if (u <= 0.0) return Double.NEGATIVE_INFINITY;
return Math.log(u / (1.0 - u)) / lambda + alpha;
}
public double evaluate(double s) {
if (s <= 0) return 1.0e100;
double sum = 0.0;
double p = s / (s + mean);
for (int j = 0; j < max; j++) sum += Fj[j] / (s + (double) j);
return sum + m * Math.log(p);
}
/** Computes the distribution function. */
public static double cdf(double n, double p, int x) {
final double EPSILON = DiscreteDistributionInt.EPSILON;
final int LIM1 = 100000;
double sum, term, termmode;
int i, mode;
final double q = 1.0 - p;
if (p < 0.0 || p > 1.0) throw new IllegalArgumentException("p not in [0, 1]");
if (n <= 0.0) throw new IllegalArgumentException("n <= 0.0");
if (x < 0) return 0.0;
if (p >= 1.0) // In fact, p == 1
return 1.0;
if (p <= 0.0) // In fact, p == 0
return 0.0;
// Compute the maximum term
mode = 1 + (int) Math.floor((n * q - 1.0) / p);
if (mode < 0) mode = 0;
else if (mode > x) mode = x;
if (mode <= LIM1) {
sum = term = termmode = prob(n, p, mode);
for (i = mode; i > 0; i--) {
term *= i / (q * (n + i - 1.0));
if (term < EPSILON) break;
sum += term;
}
term = termmode;
for (i = mode; i < x; i++) {
term *= q * (n + i) / (i + 1);
if (term < EPSILON) break;
sum += term;
}
if (sum <= 1.0) return sum;
else return 1.0;
} else
// return 1.0 - BinomialDist.cdf (x + n, p, n - 1);
return BetaDist.cdf(n, x + 1.0, 15, p);
}
public void fcn(int m, int n, double[] x, double[] fvec, double[][] fjac, int iflag[]) {
if (x[2] <= 0.0) {
final double BIG = 1.0e100;
fvec[1] = BIG;
fvec[2] = BIG;
fjac[1][1] = BIG;
fjac[1][2] = 0.0;
fjac[2][1] = 0.0;
fjac[2][2] = BIG;
return;
}
double sum;
double prod;
if (iflag[1] == 1) {
sum = 0.0;
for (int i = 0; i < n; i++) sum += (1.0 / (1.0 + Math.exp(x[2] * (xi[i] - x[1]))));
fvec[1] = sum - n / 2.0;
sum = 0.0;
for (int i = 0; i < n; i++) {
prod = x[2] * (xi[i] - x[1]);
sum -= prod * Math.tanh(prod / 2.0);
}
fvec[2] = sum - n;
} else if (iflag[1] == 2) {
sum = 0.0;
for (int i = 0; i < n; i++) {
prod = Math.exp(x[2] * (xi[i] - x[1]));
sum -= x[2] * prod / ((1 + prod) * (1 + prod));
}
fjac[1][1] = sum;
sum = 0.0;
for (int i = 0; i < n; i++) {
prod = Math.exp(x[2] * (xi[i] - x[1]));
sum -= (xi[i] - x[1]) * prod / ((1 + prod) * (1 + prod));
}
fjac[1][2] = sum;
sum = 0.0;
for (int i = 0; i < n; i++) {
prod = Math.exp(x[2] * (xi[i] - x[1]));
sum -=
(x[2] * ((-1.0 + prod) * (1.0 + prod) - (2.0 * (x[2] * (xi[i] - x[1])) * prod)))
/ ((1.0 + prod) * (1.0 + prod));
}
fjac[2][1] = sum;
sum = 0.0;
for (int i = 0; i < n; i++) {
prod = Math.exp(x[2] * (xi[i] - x[1]));
sum -=
((x[1] - xi[1])
* ((-1.0 + prod) * (1.0 + prod) - (2.0 * (x[2] * (xi[i] - x[1])) * prod)))
/ ((1.0 + prod) * (1.0 + prod));
}
fjac[2][2] = sum;
}
}
/**
* Computes and returns the standard deviation of the logistic distribution with
* parameters @f$\alpha@f$ and @f$\lambda@f$.
*
* @return the standard deviation of the logistic distribution
*/
public static double getStandardDeviation(double alpha, double lambda) {
if (lambda <= 0.0) throw new IllegalArgumentException("lambda <= 0");
return (Math.sqrt(1.0 / 3.0) * Math.PI / lambda);
}
/** Computes the complementary distribution function @f$1-F(x)@f$. */
public static double barF(double alpha, double lambda, double x) {
if (lambda <= 0) throw new IllegalArgumentException("lambda <= 0");
double z = lambda * (x - alpha);
if (z <= 100.0) return 1.0 / (1.0 + Math.exp(z));
return Math.exp(-z);
}
/**
* Sets the parameter <SPAN CLASS="MATH"><I>n</I></SPAN> and <SPAN CLASS="MATH"><I>p</I></SPAN> of
* this object.
*/
public void setParams(double n, double p) {
/* *
* Compute all probability terms of the negative binomial distribution;
* start at the mode, and calculate probabilities on each side until they
* become smaller than EPSILON. Set all others to 0.
*/
supportA = 0;
int i, mode, Nmax;
int imin, imax;
double sum;
double[] P; // Negative Binomial mass probabilities
double[] F; // Negative Binomial cumulative
if (p < 0.0 || p > 1.0) throw new IllegalArgumentException("p not in [0, 1]");
if (n <= 0.0) throw new IllegalArgumentException("n <= 0");
this.n = n;
this.p = p;
// Compute the mode (at the maximum term)
mode = 1 + (int) Math.floor((n * (1.0 - p) - 1.0) / p);
/* *
For mode > MAXN, we shall not use pre-computed arrays.
mode < 0 should be impossible, unless overflow of long occur, in
which case mode will be = LONG_MIN.
*/
if (mode < 0.0 || mode > MAXN) {
pdf = null;
cdf = null;
return;
}
/* *
In theory, the negative binomial distribution has an infinite range.
But for i > Nmax, probabilities should be extremely small.
Nmax = Mean + 16 * Standard deviation.
*/
Nmax = (int) (n * (1.0 - p) / p + 16 * Math.sqrt(n * (1.0 - p) / (p * p)));
if (Nmax < 32) Nmax = 32;
P = new double[1 + Nmax];
double epsilon = EPSILON / prob(n, p, mode);
// We shall normalize by explicitly summing all terms >= epsilon
sum = P[mode] = 1.0;
// Start from the maximum and compute terms > epsilon on each side.
i = mode;
while (i > 0 && P[i] >= epsilon) {
P[i - 1] = P[i] * i / ((1.0 - p) * (n + i - 1));
i--;
sum += P[i];
}
imin = i;
i = mode;
while (P[i] >= epsilon) {
P[i + 1] = P[i] * (1.0 - p) * (n + i) / (i + 1);
i++;
sum += P[i];
if (i == Nmax - 1) {
Nmax *= 2;
double[] nT = new double[1 + Nmax];
System.arraycopy(P, 0, nT, 0, P.length);
P = nT;
}
}
imax = i;
// Renormalize the sum of probabilities to 1
for (i = imin; i <= imax; i++) P[i] /= sum;
// Compute the cumulative probabilities for F and keep them in the
// lower part of CDF.
F = new double[1 + Nmax];
F[imin] = P[imin];
i = imin;
while (i < imax && F[i] < 0.5) {
i++;
F[i] = F[i - 1] + P[i];
}
// This is the boundary between F (i <= xmed) and 1 - F (i > xmed) in
// the array CDF
xmed = i;
// Compute the cumulative probabilities of the complementary
// distribution 1 - F and keep them in the upper part of the array
F[imax] = P[imax];
i = imax - 1;
do {
F[i] = P[i] + F[i + 1];
i--;
} while (i > xmed);
xmin = imin;
xmax = imax;
pdf = new double[imax + 1 - imin];
cdf = new double[imax + 1 - imin];
System.arraycopy(P, imin, pdf, 0, imax + 1 - imin);
System.arraycopy(F, imin, cdf, 0, imax + 1 - imin);
}
/**
* Computes and returns the standard deviation of the negative binomial distribution with
* parameters <SPAN CLASS="MATH"><I>n</I></SPAN> and <SPAN CLASS="MATH"><I>p</I></SPAN>.
*
* @return the standard deviation of the negative binomial distribution
*/
public static double getStandardDeviation(double n, double p) {
return Math.sqrt(NegativeBinomialDist.getVariance(n, p));
}
/** Computes the inverse function without precomputing tables. */
public static int inverseF(double n, double p, double u) {
if (u < 0.0 || u > 1.0) throw new IllegalArgumentException("u is not in [0,1]");
if (p < 0.0 || p > 1.0) throw new IllegalArgumentException("p not in [0, 1]");
if (n <= 0.0) throw new IllegalArgumentException("n <= 0");
if (p >= 1.0) // In fact, p == 1
return 0;
if (p <= 0.0) // In fact, p == 0
return 0;
if (u <= prob(n, p, 0)) return 0;
if (u >= 1.0) return Integer.MAX_VALUE;
double sum, term, termmode;
final double q = 1.0 - p;
// Compute the maximum term
int mode = 1 + (int) Math.floor((n * q - 1.0) / p);
if (mode < 0) mode = 0;
int i = mode;
term = prob(n, p, i);
while ((term >= u) && (term > Double.MIN_NORMAL)) {
i /= 2;
term = prob(n, p, i);
}
if (term <= Double.MIN_NORMAL) {
i *= 2;
term = prob(n, p, i);
while (term >= u && (term > Double.MIN_NORMAL)) {
term *= i / (q * (n + i - 1.0));
i--;
}
}
mode = i;
sum = termmode = prob(n, p, i);
for (i = mode; i > 0; i--) {
term *= i / (q * (n + i - 1.0));
if (term < EPSILON) break;
sum += term;
}
term = termmode;
i = mode;
double prev = -1;
if (sum < u) {
// The CDF at the mode is less than u, so we add term to get >= u.
while ((sum < u) && (sum > prev)) {
term *= q * (n + i) / (i + 1);
prev = sum;
sum += term;
i++;
}
} else {
// The computed CDF is too big so we substract from it.
sum -= term;
while (sum >= u) {
term *= i / (q * (n + i - 1.0));
i--;
sum -= term;
}
}
return i;
}