This is a proposed solution to #197. See the bullet point "Lower bound on the simulated values" in this section of the specifications and the related bullet points on obs_do_sim_constraint and obs_sim_lower_bound in this section.
thinking about NaNs from negative binomial densities in particular. should do the same thing we do for hazard function thresholding with a user-exposed option
from JD: I like x + eps*exp(-x/eps) as a posfun. Just made it up, but I'd be surprised if it hasn't been thought of before.
Specs
This is a proposed solution to #197. See the bullet point "Lower bound on the simulated values" in this section of the specifications and the related bullet points on
obs_do_sim_constraint
andobs_sim_lower_bound
in this section.Notes
https://github.com/bbolker/bbmisc/blob/master/posfun.pdf
thinking about NaNs from negative binomial densities in particular. should do the same thing we do for hazard function thresholding with a user-exposed option
from JD: I like x + eps*exp(-x/eps) as a posfun. Just made it up, but I'd be surprised if it hasn't been thought of before.