Skip to contents

Share of each haul or cruise in the annual summed density

Source: R/influence.R
density_share.Rd

A cheap, model-free screen for hauls, or groups of hauls such as cruises, that may dominate an annual index value. For each year the swept-area density of each haul (value_col / effort_col) is divided by the sum of densities over all hauls of that year. With group_col, densities are first summed within each group, so the share is the group's share of the year's summed density.

Use it to choose which identifiers to refit with sdmTMB_leave_out(): leave-one-out refits are expensive, and the identifiers with the largest density shares are the only ones that can move an annual index much. A screen of the two largest shares per year plus every share of at least 5 % (hauls) or 10 % (cruises) has worked well.

Usage

density_share(
  data,
  value_col,
  effort_col = "swept_area",
  time_col = "year",
  group_col = NULL,
  top_n = NULL,
  min_share = NULL
)

Arguments

data

A data frame with one row per haul.

value_col

Name of the catch column (biomass or numbers).

effort_col

Name of the effort column (e.g. swept area).

time_col

Name of the year column. Default "year".

group_col

Optional name of a grouping column (e.g. a cruise identifier). If NULL (default), shares are calculated per haul. Cruise numbers are often reused by different vessels, so build the group from the platform and the cruise number together.

top_n, min_share

Optional screen: candidate is TRUE for the top_n largest shares of each year and for every share of at least min_share. Both NULL (default) gives no candidate column.

Value

A tibble. Without group_col, data with the columns density, share, share_rank (1 = largest share of the year) and haul_share (1 / number of hauls in the year) added. With group_col, one row per group and year with n_hauls, n_positive, density (summed), share, share_rank and haul_share (the group's share of the year's hauls). share / haul_share is a model-free analogue of the leverage returned by sdmTMB_leave_out(). A candidate column is added when top_n or min_share is given.

See also

sdmTMB_leave_out(), index_by_region()

Author

Mikko Vihtakari

Examples

data(sebastes)
# Haul shares, screened to the two largest per year and any share >= 5 %
hs <- density_share(sebastes, value_col = "biomass",
                    effort_col = "swept_area", top_n = 2, min_share = 0.05)
sum(hs$candidate)
#> [1] 47
# Survey-series shares (with cruise data, group by platform + cruise)
ss <- density_share(sebastes, value_col = "biomass",
                    effort_col = "swept_area", group_col = "cruiseseries")
head(ss[order(-ss$share / ss$haul_share), ])
#> # A tibble: 6 × 8
#>   cruiseseries  year n_hauls n_positive density share share_rank haul_share
#>   <chr>        <int>   <int>      <int>   <dbl> <dbl>      <int>      <dbl>
#> 1 EggaS         2014      47         26 117421. 0.192          3     0.0680
#> 2 WinterS       2008     196        149 327937. 0.610          1     0.257 
#> 3 WinterS       2017     260        177 834558. 0.789          1     0.361 
#> 4 EggaS         2009      41         30  59912. 0.113          3     0.0543
#> 5 WinterS       2004     185        113 131597. 0.470          1     0.234 
#> 6 EggaS         2012      42         31  67728. 0.125          3     0.0632