Process residuals for an Rceattle model's random-effect processes

One-sample process residuals for the model's random-effect (or penalized) process deviations, in the style of SAM's procres() (Nielsen and Berg 2014). They validate the process model – whether the deviations behave like their assumed iid normal process – a complement to the observation-based osa_residuals().

Each set of deviations carries a Gaussian process prior in the model. The posterior mode of the deviations is shrunk toward that prior, so – following SAM – a single draw is taken from the joint posterior of the deviations (from the joint precision when they are random effects, or the fixed-effect covariance when they are penalized fixed effects) and standardized by the process standard deviation. Under a correctly specified process these are approximately iid N(0, 1). Because a random draw is used the residuals are stochastic; set seed for reproducibility.

Supported processes and their deviations:

"recruitment"

rec_dev, prior N(-bias_adjust_proc * R_sd^2/2, R_sd^2) per species.

"initial"

init_dev, prior N(-bias_adjust_proc * R_sd^2/2, R_sd^2) per species.

"catchability"

index_q_dev, prior N(0, q_dev_sd^2) per index.

"all" returns every supported process present in the fit. Selectivity and natural-mortality deviations (which use random-walk / 2D-AR1 priors) are not yet supported. Recruitment and initial-abundance residuals are exact (their priors are iid). Catchability residuals are exact only for the iid deviate prior (Time_varying_q = 1 or 2); for a random-walk or AR1 catchability prior the marginal-SD standardization ignores the prior correlation, so those residuals are approximate and a warning is emitted.

Usage

process_residuals(
  fit,
  process = c("recruitment", "initial", "catchability", "all"),
  seed = 123
)

Arguments

fit

A fitted Rceattle model. The targeted deviations must be estimated – as random effects (e.g. random_rec = TRUE) or as penalized fixed effects – with a usable covariance.

process

One of "recruitment", "initial", "catchability", or "all".

seed

Seed for the posterior draw. Default 123.

Value

A data frame of class rceattle_osa (so it can be passed to osa_diagnostics() and plot.rceattle_osa()) with one row per deviation: columns source (the process), fleet, species, year, age_length_bin, and residual.

References

Nielsen, A., and Berg, C.W. 2014. Estimation of time-varying selectivity in stock assessments using state-space models. Fish. Res. 158:96-101.

See also

osa_residuals(), osa_diagnostics()