cschwarz-stat-sfu-ca
commented
5 years ago Now deprecated. If you specify anything other than sampfrac=rep(1, length(u2)), the corresponding routine will terminate with an error.
In a later release we should remove all code refering to sampfrac
The sampling fraction parameter was an idea to adjust u2 (but not m2) for the length of time that the trap was operating. For example, suppose that the strata were 8 days but the trap was only operating for 4 of the 8 days. Then u2 would internally be multipled by 2 to account for this sampling fraction assuming that the capture probabilities were still valid.
I don't think this is working properly and should likely be discarded. If the user wants to adjust the data to account for traps not operating the full stratum, they can still do this outside of the models.
There is also a bias introduced when traps only run for the first half fo the stratum. Some fish released near the end of the stratum have NO chance of being selected. See the Taku river vignette for an example of potential bias being introduced.
I think the solution for this problem is to create a large set of smaller strata with some p's being set to zero.