mmmstan: Fit A Hidden Markov Movement Model via CmdStan and CmdStanR
In luke-a-rogers/mmmstan: Fit Markov Movement Models (MMM) via CmdStan and CmdStanR

mmmstan

R Documentation

Fit A Hidden Markov Movement Model via CmdStan and CmdStanR

Description

The Stan model data may be specified by individual arguments or by passing a suitable data argument. The data argument takes precedence when it is non-null.

Usage

mmmstan(
  tag_data,
  list_regions,
  list_sizes,
  year_start,
  year_end,
  step_interval = "quarter",
  step_duration_max = NULL,
  days_duration_min = 90,
  colname_date_released = "date_released",
  colname_date_recovered = "date_recovered",
  colname_region_released = "region_released",
  colname_region_recovered = "region_recovered",
  colname_size_released = "size_released",
  movement_pattern = 2,
  movement_allow = NULL,
  movement_disallow = NULL,
  mu_movement_step_diag = NULL,
  sd_movement_step_diag = NULL,
  mu_fishing_rate = NULL,
  cv_fishing_rate = NULL,
  mu_selectivity = NULL,
  cv_selectivity = NULL,
  mu_fishing_weight = NULL,
  sd_fishing_weight = NULL,
  mu_natural_mortality_rate = NULL,
  sd_natural_mortality_rate = NULL,
  mu_reporting_rate = NULL,
  sd_reporting_rate = NULL,
  mu_initial_loss_rate = 0.1,
  sd_initial_loss_rate = 0.01,
  mu_ongoing_loss_rate = 0.02,
  sd_ongoing_loss_rate = 0.001,
  mu_dispersion = 1,
  sd_dispersion = 0.5,
  tolerance_expected = 1e-12,
  tolerance_fishing = 1e-12,
  data = NULL,
  chains = 1,
  step_size = 0.01,
  adapt_delta = 0.95,
  iter_warmup = 250,
  iter_sampling = 750,
  max_treedepth = 10,
  use_reduce_sum = FALSE,
  threads_per_chain = parallel::detectCores()/(2 * chains),
  refresh = 100,
  ...
)

Arguments

`tag_data`	`data.frame()`
`list_regions`	`list()`
`list_sizes`	`list()`
`year_start`	`integer()` year of initial tag released
`year_end`	`integer()` year of final tag recovered
`step_interval`	`character()` one of `"month", "quarter", "year"`
`step_duration_max`	`integer()`
`days_duration_min`	`integer()` minimum number of days before recovery
`colname_date_released`	`character()`
`colname_date_recovered`	`character()`
`colname_region_released`	`character()`
`colname_region_recovered`	`character()`
`colname_size_released`	`character()`
`movement_pattern`	`integer()`
`movement_allow`	integer()
`movement_disallow`	integer()
`mu_movement_step_diag`	numeric()
`sd_movement_step_diag`	numeric()
`mu_fishing_rate`	numeric()
`cv_fishing_rate`	`numeric()`
`mu_selectivity`	numeric()
`cv_selectivity`	`numeric()`
`mu_fishing_weight`	`numeric()`
`sd_fishing_weight`	`numeric()`
`mu_natural_mortality_rate`	`numeric()`
`sd_natural_mortality_rate`	`numeric()`
`mu_reporting_rate`	`numeric()`
`sd_reporting_rate`	`numeric()`
`mu_initial_loss_rate`	`numeric()`
`sd_initial_loss_rate`	`numeric()`
`mu_ongoing_loss_rate`	`numeric()`
`sd_ongoing_loss_rate`	`numeric()`
`mu_dispersion`	`numeric()`
`sd_dispersion`	`numeric()`
`tolerance_expected`	`numeric()`
`tolerance_fishing`	`numeric()`
`data`	`list()` See details
`chains`	`integer()` number of chains
`step_size`	`numeric()` initial step size
`adapt_delta`	`numeric()` the adaptation target acceptance statistic
`iter_warmup`	`integer()` number of warmup iterations
`iter_sampling`	`integer()` number of sampling iterations
`max_treedepth`	`integer()`
`use_reduce_sum`	`logical()` use within chain parallel threading
`threads_per_chain`	`integer()` number of threads per chain
`refresh`	`integer()`
`...`	additional arguments to pass to `$sample()` method

Details

When the model data are specified by individual arguments, a number of helper functions with sensible defaults are used to assemble the data object. This is usually preferable for user convenience.

Passing the data object directly may be preferred for example when the data object is the output of a simulation model, or when greater flexibility is desired. In either event, care must be taken to ensure the data object matches the requirements of the underlying Stan model.

TBD