All model configurations require the initialization of the age-specific densities at the first year (). This initialization applies to all age classes greater than one (), as the density of the first age-class is always determined by the estimated recruitment.
The drmr package provides three methods for setting these initial values, each controlled by a specific parameter or toggle in the fit_drm function. The sections below detail each of these options.
User provided densities
This is the most direct method, activated by providing a numeric vector to the init_data argument in fit_drm.
How it works: The user supplies a vector of log-transformed densities of length
n_ages - 1, corresponding to the initial densities for ages2throughn_ages(total number of age classes).Limitation: This method applies the same initial age structure to all patches, which may not capture pre-existing spatial variation.
Estimate age-specific initial densities
This option is enabled by setting the est_init = 1 toggle.
How it works: Instead of requiring fixed inputs, the model treats the initial log-densities for ages
2throughn_agesas free parameters to be estimated.Limitation: Similar to the user-provided option, this method estimates a single set of initial densities that are applied uniformly across all patches. We also caution that this approach can introduce substantial uncertainty, sometimes resulting in very wide prediction intervals.
Estimate age-specific densities from mortality rates
This dynamic approach is activated by setting the minit = 1 toggle. It is particularly useful when external, age-specific mortality data is available.
- How it works: This method assumes the population was at a stable age distribution just before the first year of observation. It uses the model’s estimated (or externally provided) mortality and recruitment rates to reconstruct the initial age-specific densities for each patch.
Unlike the other methods, this allows the initial age structure to vary spatially across patches, reflecting local differences in estimated recruitment and survival rates.