# -*- org -*-
#+CATEGORY: cdf

* discrete inverse distributions - easy apart from hypergeometric(?)

* look for integer overflow in the discrete functions - this could be
  hard to find - perform computations in double to avoid.

    gsl_cdf_binomial_P (unsigned int k, double p, unsigned int n);
    gsl_cdf_binomial_Q (unsigned int k, double p, unsigned int n);
    gsl_cdf_poisson_P (unsigned int k, double mu);
    gsl_cdf_poisson_Q (unsigned int k, double mu);
    gsl_cdf_geometric_P (unsigned int k, double p);
    gsl_cdf_geometric_Q (unsigned int k, double p);
    gsl_cdf_negative_binomial_P (unsigned int k, double p, double n);
    gsl_cdf_negative_binomial_Q (unsigned int k, double p, double n);
    gsl_cdf_pascal_P (unsigned int k, double p, unsigned int n);
    gsl_cdf_pascal_Q (unsigned int k, double p, unsigned int n);
    gsl_cdf_hypergeometric_P (unsigned int k, unsigned int n1,
                              unsigned int n2, unsigned int t);
    gsl_cdf_hypergeometric_Q (unsigned int k, unsigned int n1,
                              unsigned int n2, unsigned int t);

* Unify the beta_inc function with the special functions, put all the
  functionaity in gsl_sf_beta_inc, providing a new function for the
  AXPY part if necessary (can we do everything with gsl_sf_beta_inc
  and gsl_sf_beta_inc_1mx keeping in mind that we cannot do 1-x
  because of cancellation for small x)