Starting Values and Fixing (and Sharing) Parameters

In addition to the SPoRC::Setup_xxx functions, users can access several advanced model features. These include:

  • Specifying starting values (closely tied to parameter fixing)
  • Fixing parameters
  • Sharing parameters across model partitions

Starting Values

Starting values can be specified in two ways:

  1. Directly within the SPoRC::Setup_xxx functions. Each setup function accepts starting values through the ... argument. The inputs must match the model’s parameter names and dimensions. For more information on parameter dimensions, see the Description of Model Parameters vignette.

  • SPoRC::Setup_Mod_Rec
    ln_global_R0, Rec_prop, steepness_h, ln_InitDevs, ln_RecDevs, ln_sigmaR

  • SPoRC::Setup_Mod_Biologicals
    ln_M, M_offset

  • SPoRC::Setup_Mod_Movement
    move_pars, move_devs, move_pe_pars, log_move_diffusion_pars, move_preference_pars

  • SPoRC::Setup_Mod_Tagging
    ln_Init_Tag_Mort, ln_Tag_Shed, ln_tag_theta, Tag_Reporting_Pars

  • SPoRC::Setup_Mod_Catch_and_F
    ln_sigmaC, ln_sigmaF, ln_sigmaF_agg, ln_F_mean, ln_F_devs, ln_F_mean_AggCatch, ln_F_devs_AggCatch

  • SPoRC::Setup_Mod_FishIdx_and_Comps
    ln_FishAge_theta, FishAge_corr_pars, ln_FishAge_theta_agg, FishAge_corr_pars_agg,
    ln_FishLen_theta, FishLen_corr_pars, ln_FishLen_theta_agg, FishLen_corr_pars_agg

  • SPoRC::Setup_Mod_SrvIdx_and_Comps
    ln_SrvAge_theta, SrvAge_corr_pars, ln_SrvAge_theta_agg, SrvAge_corr_pars_agg,
    ln_SrvLen_theta, SrvLen_corr_pars, ln_SrvLen_theta_agg, SrvLen_corr_pars_agg

  • SPoRC::Setup_Mod_Fishsel_and_Q
    ln_fish_fixed_sel_pars, ln_fish_q, fishsel_pe_pars, ln_fishsel_devs

  • SPoRC::Setup_Mod_Srvsel_and_Q
    ln_srv_fixed_sel_pars, ln_srv_q, srvsel_pe_pars, ln_srvsel_devs

  1. Post-hoc modification of starting values. Alternatively, you can first call the setup functions without specifying starting values, then access and modify the internally created parameter list (input_list$par) before running the model.

In the following, we will illustrate both methods using the recruitment module (Setup_Mod_Rec) and specify starting values for ln_global_R0 and ln_sigmaR. First let us load in the package and define the model dimensions.

# Load in packages
library(SPoRC) 
data("sgl_rg_sable_data") # load in data

input_list <- Setup_Mod_Dim(years = 1:length(sgl_rg_sable_data$years), # vector of years 
                            # (corresponds to year 1960 - 2024)
                            ages = 1:length(sgl_rg_sable_data$ages), # vector of ages
                            lens = seq(41,99,2), # number of lengths
                            n_regions = 1, # number of regions
                            n_sexes = sgl_rg_sable_data$n_sexes, # number of sexes == 1,
                            # female, == 2 male
                            n_fish_fleets = sgl_rg_sable_data$n_fish_fleets, # number of fishery
                            # fleet == 1, fixed gear, == 2 trawl gear
                            n_srv_fleets = sgl_rg_sable_data$n_srv_fleets, # number of survey fleets
                            verbose = FALSE
                            )

We can specify starting values directly using the Setup_Mod_Rec function. Note that all inputs passed via the ... argument must exactly match the parameter names and their expected dimensions in the model (see the Description of Model Parameters vignette for details).

input_list <- Setup_Mod_Rec(
  input_list = input_list,        # input data list from above
  # Model options
  do_rec_bias_ramp = 0,       # disable bias ramp
  sigmaR_switch = as.integer(length(1960:1975)),  # switch from early to late sigmaR
  dont_est_recdev_last = 1,       # do not estimate last recruitment deviate
  rec_model = "mean_rec",         # recruitment model type
  init_age_strc = 1,              # geometric series for initial age structure
  
  # Specify starting values
  ln_global_R0 = log(30),         # starting value for global R0
  ln_sigmaR = c(log(1.5), log(1.5))  # starting values for early and late sigmaR
)

In this example, the starting value for ln_global_R0 is set to log(30), while ln_sigmaR is set to log(1.5) for both the early (first element) and late (second element) periods.

Alternatively, starting values can be assigned after running the setup functions without initial specifications. Users can extract the internal parameter list and modify starting values as needed:

input_list <- Setup_Mod_Rec(input_list = input_list, # input data list from above
                            # Model options
                            do_rec_bias_ramp = 0, # don't do bias ramp 
                            sigmaR_switch = as.integer(length(1960:1975)), # when to switch from early to late sigmaR
                            dont_est_recdev_last = 1, # don't estimate last recruitment deviate
                            rec_model = "mean_rec", # recruitment model
                            init_age_strc = 1 # geometric series to derive age structure
                            )

# Specify starting values post-hoc
# R0
input_list$par$ln_global_R0 # default starting value
#> [1] 2.70805
input_list$par$ln_global_R0 <- log(30) # user specified starting value

# sigmaR
input_list$par$ln_sigmaR # default starting value
#> [1] 0 0
input_list$par$ln_sigmaR[] <- c(log(1.5), log(1.5)) # user specified starting value

Mapping

Mapping is a core feature of TMB and RTMB models. It allows users to either fix parameters at known values or to share parameters across different parts of the model. In the sections below, we first demonstrate how to use mapping to fix parameters. We then show how mapping can be used to share parameters across model partitions.

Fixing Parameters

The SPoRC::Setup_xxx functions include arguments that allow certain parameters to be fixed, meaning they are not estimated during model fitting. These arguments are not available for all parameters, but we note that all parameters can be specified to be fixed. As an example, we use Setup_Mod_Rec to show how the ln_sigmaR parameter can be fixed. The argument sigmaR_spec = "fix".

input_list <- Setup_Mod_Rec(input_list = input_list, # input data list from above
                            # Model options
                            do_rec_bias_ramp = 0, # don't do bias ramp 
                            sigmaR_switch = as.integer(length(1960:1975)), # when to switch from early to late sigmaR
                            dont_est_recdev_last = 1, # don't estimate last recruitment deviate
                            rec_model = "mean_rec", # recruitment model
                            init_age_strc = 1, # geometric series to derive age structure
                            
                            # Parameter Fixing
                            sigmaR_spec = 'fix'
                            )

input_list$map$ln_sigmaR # both values are fixed and not estimated (specified as factor(rep(NA, 2)))
#> [1] <NA> <NA>
#> Levels:
input_list$par$ln_sigmaR # ln_sigmaR is then fixed at the default starting value
#> [1] 0 0

To fix ln_sigmaR at a specific value, simply supply that value when calling the function:

input_list <- Setup_Mod_Rec(input_list = input_list, # input data list from above
                            # Model options
                            do_rec_bias_ramp = , # don't do bias ramp 
                            sigmaR_switch = as.integer(length(1960:1975)), # when to switch from early to late sigmaR
                            dont_est_recdev_last = 1, # don't estimate last recruitment deviate
                            rec_model = "mean_rec", # recruitment model
                            init_age_strc = 1, # geometric series to derive age structure
                            
                            # Parameter Fixing
                            sigmaR_spec = 'fix',
                            ln_sigmaR = c(log(1.5), log(1.5)) # user specified starting value
                            )

input_list$map$ln_sigmaR # both values are fixed and not estimated (specified as factor(rep(NA, 2)))
#> [1] <NA> <NA>
#> Levels:
input_list$par$ln_sigmaR # ln_sigmaR is then fixed at the user specified starting value
#> [1] 0.4054651 0.4054651

However, not all parameters include a convenience argument like sigmaR_spec = "fix". For example, ln_global_R0 does not. In such cases, you can fix the parameter manually by modifying the map list directly and specifying the desired starting value:

input_list <- Setup_Mod_Rec(input_list = input_list, # input data list from above
                            # Model options
                            do_rec_bias_ramp = 0, # don't do bias ramp 
                            sigmaR_switch = as.integer(length(1960:1975)), # when to switch from early to late sigmaR
                            dont_est_recdev_last = 1, # don't estimate last recruitment deviate
                            rec_model = "mean_rec", # recruitment model
                            init_age_strc = 1, # geometric series to derive age structure
                            )

input_list$map$ln_global_R0 <- factor(NA)
input_list$par$ln_global_R0 <- log(30)

In this example, ln_global_R0 is fixed at log(30) by setting its map entry to NA and providing the desired value in the parameter list.

Sharing Parameters

The SPoRC::Setup_xxx functions also support sharing parameters across model partitions. For example, selectivity parameters can be shared between sexes using built-in convenience arguments. These options simplify common sharing structures, but users can also implement more customized behavior when needed. Below, we demonstrate two approaches. First, we show how to share fishery selectivity parameters between sexes using a convenience flag in Setup_Mod_Fishsel_and_Q. Then, we outline how users can manually configure parameter sharing for finer control.

In the example below, we define two fishery fleets. Fleet 1 uses a logistic selectivity model (2 parameters), while Fleet 2 uses a gamma dome-shaped model (2 parameters). We first use the fish_fixed_sel_pars_spec argument to specify that selectivity parameters for both fleets should be estimated across all model partitions (est_all). Thus, when we inspect the map, we should expect a total of 8 unique numbers, which represent unique selectivity parameters to be estimated for each sex and fleet (4 for each sex and fleet combination).

input_list$data$Selex_Type <- 0 # specifying age-based selectivity for demonstration purposes
input_list <- SPoRC::Setup_Mod_Fishsel_and_Q(input_list = input_list,

                                      # Model options
                                      cont_tv_fish_sel = c("none_Fleet_1", "none_Fleet_2"),
                                      fish_sel_blocks = c("none_Fleet_1", "none_Fleet_2"),
                                      fish_sel_model = c("logist1_Fleet_1", "gamma_Fleet_2"),
                                      fish_q_blocks = c("none_Fleet_1", "none_Fleet_2"),
                                      fish_q_spec = c("fix", "fix"),

                                      # Share selectivity parameters across all partitions
                                      fish_fixed_sel_pars_spec = c("est_all", "est_all"))

input_list$map$ln_fish_fixed_sel_pars # 8 unique numbers, 4 for each sex and fleet combination
#> [1] <NA> <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> Levels:

Next, we use the fish_fixed_sel_pars_spec argument to indicate that selectivity parameters should be shared across sex partitions for both fleets. Specifically, we use the "est_shared_s" setting to link parameters between sexes, even though the model is sex-structured. In this setup, we still estimate separate parameters for each fleet, but the same selectivity parameters are used for both sexes within each fleet. As a result, when we inspect the mapping, we should expect four unique values—two per fleet—despite having two sexes.

input_list$data$Selex_Type <- 0 # specifying age-based selectivity for demonstration purposes
input_list <- SPoRC::Setup_Mod_Fishsel_and_Q(input_list = input_list,

                                      # Model options
                                      cont_tv_fish_sel = c("none_Fleet_1", "none_Fleet_2"),
                                      fish_sel_blocks = c("none_Fleet_1", "none_Fleet_2"),
                                      fish_sel_model = c("logist1_Fleet_1", "gamma_Fleet_2"),
                                      fish_q_blocks = c("none_Fleet_1", "none_Fleet_2"),
                                      fish_q_spec = c("fix", "fix"),

                                      # Share selectivity parameters across all partitions
                                      fish_fixed_sel_pars_spec = c("est_shared_s", "est_shared_s"))

input_list$map$ln_fish_fixed_sel_pars # 4 unique numbers, 2 for each sex and fleet combination
#> [1] <NA> <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> Levels:

The parameter array ln_fish_fixed_sel_pars has the following structure: [n_regions, n_max_sel_pars, n_max_sel_blocks, n_sexes, n_fish_fleets] In this example, its dimensions are 1 x 2 x 1 x 2 x 2.

This means that:

  • For each fleet, there are two estimated selectivity parameters

  • These are shared across the two sexes

  • When flattened into a vector (as required by the mapping), this results in four distinct values (two per fleet), even though the underlying array spans two sexes

Below is a minimal example that replicates what Setup_Mod_Fishsel_and_Q is doing internally:

# define dimensions
n_regions <- 1
n_max_sel_pars <- 2
n_max_sel_blocks <- 1
n_sexes <- 2
n_fish_fleets <- 2

# define empty parameter array for demonstration
ln_fish_fixed_sel_pars <- array(0, dim = c(n_regions,n_max_sel_pars,n_max_sel_blocks,n_sexes,n_fish_fleets))
ln_fish_fixed_sel_pars[1,,1,,1] <- c(1,2) # same parameters for each sex for fleet 1
ln_fish_fixed_sel_pars[1,,1,,2] <- c(3,4) # same parameters for each sex for fleet 2
ln_fish_fixed_sel_pars
#> , , 1, 1, 1
#> 
#>      [,1] [,2]
#> [1,]    1    2
#> 
#> , , 1, 2, 1
#> 
#>      [,1] [,2]
#> [1,]    1    2
#> 
#> , , 1, 1, 2
#> 
#>      [,1] [,2]
#> [1,]    3    4
#> 
#> , , 1, 2, 2
#> 
#>      [,1] [,2]
#> [1,]    3    4

# flatten array to a vector of factors for the map
custom_map <- factor(as.vector(ln_fish_fixed_sel_pars))

We can now compare the custom map object we constructed manually with the one generated by Setup_Mod_Fishsel_and_Q:

custom_map
#> [1] 1 2 1 2 3 4 3 4
#> Levels: 1 2 3 4
input_list$map$ln_fish_fixed_sel_pars
#> [1] <NA> <NA> <NA> <NA> <NA> <NA> <NA> <NA>
#> Levels:

Both should match, confirming that parameters are being correctly shared across sexes for each fleet.

Lastly, we demonstrate how to manually specify advanced mapping and parameter-sharing options for fishery selectivity parameters. In this example, we want to:

  • Estimate logistic selectivity for Fleet 1, where:
    • The \(a_{50}\) parameter (inflection point) is sex-specific
    • The slope parameter is shared across sexes
  • Estimate gamma selectivity for Fleet 2, where:
    • All parameters are sex-specific

This level of control is not supported by the built-in fish_fixed_sel_pars_spec convenience arguments, so we must construct the mapping manually.

We begin by calling Setup_Mod_Fishsel_and_Q with the "est_all" option for both fleets. This creates a parameter array with the correct dimensions and assumes that all selectivity parameters are uniquely estimated across model partitions.

input_list$data$Selex_Type <- 0 # specifying age-based selectivity for demonstration purposes
input_list <- SPoRC::Setup_Mod_Fishsel_and_Q(
  input_list = input_list,

  # Model options
  cont_tv_fish_sel = c("none_Fleet_1", "none_Fleet_2"),
  fish_sel_blocks = c("none_Fleet_1", "none_Fleet_2"),
  fish_sel_model = c("logist1_Fleet_1", "gamma_Fleet_2"),
  fish_q_blocks = c("none_Fleet_1", "none_Fleet_2"),
  fish_q_spec = c("fix", "fix"),

  # Start with all parameters estimated independently
  fish_fixed_sel_pars_spec = c("est_all", "est_all")
)

We then extract the parameter array created internally for ln_fish_fixed_sel_pars. This array has dimensions: [n_regions, n_max_sel_pars, n_max_sel_blocks, n_sexes, n_fish_fleets]. In this example, its dimensions are 1 x 2 x 1 x 2 x 2. We can then assign integer values to this array to indicate which parameters are shared (same value) or independently estimated (different values).

map_ln_fish_fixed_sel_pars <- input_list$par$ln_fish_fixed_sel_pars  # extract the array

We now assign specific values to define the desired sharing structure:

# Fleet 1 (logistic): 
# - sel_par 1 = a50 (sex-specific)
# - sel_par 2 = slope (shared across sex)
map_ln_fish_fixed_sel_pars[1,1,1,,1] <- c(1,2)   # a50: unique for each sex
map_ln_fish_fixed_sel_pars[1,2,1,,1] <- c(3)     # slope: same for both sexes

# Fleet 2 (gamma): 
# - sel_par 1 = amax (sex-specific)
# - sel_par 2 = slope (sex-specific)
map_ln_fish_fixed_sel_pars[1,1,1,,2] <- c(4,5)   # amax: unique per sex
map_ln_fish_fixed_sel_pars[1,2,1,,2] <- c(6,7)   # slope: unique per sex

We then flatten this array into a vector of factors and assign it to the map list. This tells the model how parameters should be linked when building the model.

# Flatten the mapping array and assign it to the map
input_list$map$ln_fish_fixed_sel_pars <- factor(as.vector(map_ln_fish_fixed_sel_pars))

# View the final map
input_list$map$ln_fish_fixed_sel_pars
#> [1] 1 3 2 3 4 6 5 7
#> Levels: 1 2 3 4 5 6 7

This custom mapping approach allows full control over parameter sharing structures beyond what is available through the high-level setup arguments.