Open Cole-Monnahan-NOAA opened 4 years ago
The distribution used (inverse gaussian) was for simulations based on mean and variance estimates of input recruits. It would be interesting to examine alternatives for simulations, but just to clarify there is no likelihood estimation involved (unless including an SRR in which case lognormal is used in fitting.
And to be clear, these are for simulation purposes, not exactly "likelihoods"
It may be interesting/prudent to explore the consequences of using the inverse Gaussian distribution. Grant says this has come up at the SSC before.