Returns a matrix, each row being a unit-sum observation.
Arguments
H
Object of class hyper2
n
Number of samples
startp
Starting value for the Markov chain, with default
NULL being interpreted as starting from the evaluate
fcm,fcv
Constraints as for maxp()
SMALL
Notional small value for numerical stability
l
Log-likelihood function with default loglik()
...
Further arguments, currently ignored
Author
Robin K. S. Hankin
Details
Uses the implementation of Metropolis-Hastings from the MCE
package to sample from the posterior PDF of a hyper2 object.
If the distribution is Dirichlet, use rdirichlet() to generate
random observations: it is much faster, and produces serially
independent samples. To return uniform samples, use
rp_unif() (documented at dirichlet.Rd).