function gb = gamm_rnd(nrow,m,k)
% PURPOSE: a matrix of random draws from the gamma distribution
%---------------------------------------------------
% USAGE: r = gamm_rnd(nrow,ncol,m,k)
% where: nrow = the row size of the vector drawn 
%        m = a parameter such that the mean of the gamma = m/k
%        k = a parameter (or vector) such that the variance of the gamma = m/(k^2)
%        note: m=r/2, k=2 equals chisq r random deviate 
%---------------------------------------------------
% RETURNS:
%        r = an nrow x 1 vector of random numbers from the gamma distribution      
% --------------------------------------------------
% SEE ALSO: gamm_inv, gamm_pdf, gamm_cdf
%---------------------------------------------------
% NOTE: written by: Michael Gordy, 15 Sept 1993
%                   mbgordy@athena.mit.edu
%---------------------------------------------------
% REFERENCES: Luc Devroye, Non-Uniform Random Variate Generation, 
%            New York: Springer Verlag, 1986, ch 9.3-6.

if nargin ~= 3
error('Wrong # of arguments to gamm_rnd');
end;

ncol = 1;
gb=zeros(nrow,ncol);
if m<=1
  % Use RGS algorithm by Best, p. 426
  c=1/m; 
  t=0.07+0.75*sqrt(1-m);
  b=1+exp(-t)*m/t;
  for i1=1:nrow
    for i2=1:ncol
       accept=0;
       while accept==0
          u=rand; w=rand; v=b*u;
          if v<=1
             x=t*(v^c);
             accept=((w<=((2-x)/(2+x))) | (w<=exp(-x)));
          else
             x=-log(c*t*(b-v));
             y=x/t;
             accept=(((w*(m+y-m*y))<=1) | (w<=(y^(m-1))));
          end
       end
       gb(i1,i2)=x;
    end
  end
else
  % Use Best's rejection algorithm XG, p. 410
  b=m-1;
  c=3*m-0.75;
  for i1=1:nrow
    for i2=1:ncol
       accept=0;
       while accept==0
          u=rand;  v=rand;
          w=u*(1-u);  y=sqrt(c/w)*(u-0.5);
          x=b+y;
          if x >= 0
             z=64*(w^3)*v*v;
             accept=(z<=(1-2*y*y/x)) ...
                    | (log(z)<=(2*(b*log(x/b)-y)));
          end
       end
       gb(i1,i2)=x;
    end
  end
end
gb = matdiv(gb,k);