Gadget3 CKMR likelihood

*Experimental* CKMR tagging likelihood

g3l_ckmr_po(
        nll_name,
        obs_data,
        parent_stocks,
        offspring_stocks,
        weight = g3_parameterized(paste0(nll_name, "_weight"),
            optimise = FALSE, value = 1),
        run_at = g3_action_order$likelihood)

Arguments

nll_name

Character string, used as a prefix for model history storage (nll_namemodel) and observed data variable names.

obs_data

Data.frame of observed mother-offspring pairs with columns year / parent_age / offspring_age / mo_pairs / n_comparisons

parent_stocks

A list of g3_stock objects that are parents in a g3a_spawn action

offspring_stocks

A list of g3_stock objects that are output_stocks in a g3a_spawn action

weight

Weighting applied to this likelihood component. Default is a g3_param that defaults to 1, allowing weights to be altered without recompiling.

run_at

Integer order that actions will be run within model, see g3_action_order.

Details

Implementation of CKMR based on Bravington, M.V., Skaug, H.J., & Anderson, E.C. (2016). Close-Kin Mark-Recapture. Statistical Science, 31, 259-274.

Only one kinship probability is implemented, mother-offspring with lethal sampling, i.e. (3.2) in the paper. This is then used as a pseudo-likelihood as per (4.1).

obs_data

The obs_data data.frame provides observed pairs. Unlike other likelihood methods, it has a fixed structure:

year

Year of observation for the data point.

parent_age

Age of the parent in an observed parent-offspring pair.

offspring_age

Age of the offspring in an observed parent-offspring pair.

mo_pairs

Number of mother-offspring pairs observed with these ages.

n_comparisons

Total number of comparisons made to find mo_pairs.

See also

Bravington, M.V., Skaug, H.J., & Anderson, E.C. (2016). Close-Kin Mark-Recapture. Statistical Science, 31, 259-274. g3_stock

Value

g3l_ckmr_po

An action (i.e. list of formula objects) that will...

1. During each spawning step, for all parent_stocks, accumulate total parent numbers into modelhist__num and total spawned offspring into modelhist__spawned by year and age

2. For any observed pairs in that year, compute the Poisson pseudo-likelihood of the observed mother-offspring pairs and add to nll

Examples

stocks <- list(
  mat = g3_stock(c("cod", "mat"), seq(20, 100, 10)) |> g3s_age(3, 10),
  imm = g3_stock(c("cod", "imm"), seq(5, 40, 5)) |> g3s_age(0, 2)
)

# Observed mother-offspring pairs: year / parent_age / offspring_age / mo_pairs / n_comparisons
obs_data <- data.frame(
    year          = c(2000L, 2001L),
    parent_age    = c(   6L,    7L),
    offspring_age = c(   1L,    1L),
    mo_pairs      = c(   3L,    1L),
    n_comparisons = c( 500L,  500L))

actions <- list(
    g3a_time(2000, 2002, c(6, 6), project_years = 0),
    g3a_initialconditions_normalcv(stocks$mat),
    g3a_initialconditions_normalcv(stocks$imm),
    g3a_age(stocks$mat),
    g3a_age(stocks$imm),
    g3a_spawn(
        stocks$mat,
        recruitment_f = g3a_spawn_recruitment_fecundity(
            p0 = 1, p1 = 0, p2 = 0, p3 = 1, p4 = 0),
        output_stocks = list(stocks$imm),
        run_step = 1),
    g3l_ckmr_po("ckmr", obs_data,
        parent_stocks = list(stocks$mat),
        offspring_stocks = list(stocks$imm)))