[R] Kullback–Leibler divergence question (flexmix::KLdiv) Urgent!

[Ralf B]

Hi all,

x <- cbind(rnorm(500),rnorm(500))
KLdiv(x, eps=1e-4)
KLdiv(x, eps=1e-5)
KLdiv(x, eps=1e-6)
KLdiv(x, eps=1e-7)
KLdiv(x, eps=1e-8)
KLdiv(x, eps=1e-9)
KLdiv(x, eps=1e-10)
...
KLdiv(x, eps=1e-100)
...
KLdiv(x, eps=1e-1000)

When calling flexmix::KLdiv using the given code I get results with
increasing value the smaller I pick the accuracy parameter 'eps' until
finally reaching infinite. If I pick the number too low, I get NA as a
result.

What is the best value for eps and how should one deal with this?
Should I simple pick a value that returns an result and then keep the
accuracy value constant at this level for all my analysis in order to
get comparable results?

Ralf