Handling missing records AssignPSUSamplingParameters

[edvinf]

AssignPSUSamplingParameters halts with an error if data records are not provided for all PSUs that are sampled in the PSU-sampling parameters. This is meant to ensure that correct data-sets have been obtained.

Some complications arise, however with partial measurements. An example is catchlottery-sampling of Blue Whiting in 2023. Ages have been selectively obtained, and some hauls have no age-records. So, while the PSU has indeed been sampled, it would be reasonable to assume that it is missing for purposes of estimation in age-domains.

This is currently rather cumbersome to deal with. It would be preferable to have a mechanism in StoX, for assuming these as missing at random. One suggestion would be to have an option to AssignPSUSamplingParameters which allows one to assume any missing data-records as missing at random. With such an option, unwanted records could be filtered in StoXBiotic in usual ways.

If age-less records are left in, estimation for each PSU proceeds as usual, but some PSUs will have NA estimates for some needed parameters, so that population-estimation will halt with an error. The issue could potentially be addressed also at this stage.

A StoX project encountering the issue is archived in imr internal gitlab: https://git.imr.no/a5362/samplesize-sensitivity-commercial-sampling

[edvinf]

For now, updated guidance for a simplified approach is incorporated into documentation in (#177 )

[edvinf]

This will probably be solved by removing missing serialnumbers from the sampling parameters, and issue a message or warning.