Underlying Rules

Source: vignettes/underlyingRules.Rmd

The following set of rules are the basis for mscatch.

In many stock assessments each year may result in a different level of aggregation (i.e quarters, semesters or different gear types).

Note: minimum number of length samples ~ 1 per 200 metric tonnes landed. (unwritten rule of thumb, check )

(Within each year, t)

Combine:

  • Are there gear types with total landings greater than 0 but total sample sizes less than X?
    + if YES: Combine gear types (explore to see if combined lengths similar) 
    to increase sample size or Ignore and continue (add to total at the end).
    + if NO: Count your blessings and continue.
  • Are there market categories (they are not on ordinal scale) with total landings greater than 0 but sample sizes less than X? (after combining gears).
      + if YES: combine market categories with similar length distributions. 
      + if NO : Count your blessings and continue.
  • Are there quarters with landings but sample sizes less than X?
    + if YES: Create semesters?  (Currently semesters are skipped and jump straight to annual)
      + if NO: Count your blessings and continue.
  • Are there strata with landings but sample sizes less than X?
      + If YES: define groups of similar strata  
      + If NO: Count your blessings and continue.
  • Interpolate remaining holes.
      + Select length within year/market category from adjacent strata or strata group
      (after determining if appropriate).  
      + Select length within year/market category from adjacent quarter or semester
      (after determining if appropriate).  
      + Select length within year/market category from adjacent year
      (after determining if appropriate).
  • If unclassified market category and no other information available:
      + Combine market category samples to create an unclassified  
      + Use appropriate observer length samples to apply to unclassified.
  • If holes remain
      + Calculate ratio of sampled to un-sampled landings and expand
      total length frequency such that expanded sample weight equals
      total landings weight for year t.

Interpolate only:

If examination of sample matrix shows only sporadic missing cells with landings but sample sizes less than X:

  • Samples by year/gear/quarter/stat area/market category

    + Nearest stat area within market category/quarter/gear/year  
    + Keeping year/market category constant:  
        + Adjacent quarter/ same stat area  
            + Adjacent quarter/ any stat area within ecoregion  
                + Adjacent semester/matching stat area  
                    + Adjacent semester/ any stat area within ecoregion  
                      +  If lengths by gear are similar: Alternative gear/same quarter/stat area  
                            + Alternative gear/same quarter/any stat area
                                + Alternative gear/semester/same stat area
                                    + Alternative gear/semester/any stat area
                                        + No options- extrapolate to total

Or combination of both approaches