make_osa_residuals: Calculate one-step-ahead residuals

In timjmiller/wham: Woods Hole Assessment Model (WHAM)

Calculate one-step-ahead residuals

Description

Standard residuals are not appropriate for models with random effects. Instead, one-step-ahead (OSA) residuals can be used for evaluating model goodness-of-fit (Thygeson et al. (2017), implemented in TMB::oneStepPredict). OSA residual options are passed to TMB::oneStepPredict in a list osa.opts. Current options are method: oneStepGaussianOffMode (default), oneStepGaussian, or oneStepGeneric, and parallel: TRUE/FALSE. See TMB::oneStepPredict for further details. It is not recommended to run this function (or TMB::oneStepPredict) with any random effects and mvtweedie age composition likelihoods due to extensive computational demand. An error will be thrown in such cases. See Trijoulet et al. (2023) for OSA methods for age composition OSA residuals.

Usage

make_osa_residuals(
  model,
  osa.opts = list(method = "oneStepGaussianOffMode", parallel = TRUE)
)

Arguments

model

A fit WHAM model, output from fit_wham.

Value

the same fit TMB model with additional elements for osa residuals:

$OSA.Ecov

data.frame returned by TMB::oneStepPredict for environmental observations, if applicable.

$OSA.agregate

data.frame returned by TMB::oneStepPredict for aggregate catch and index observations conditional on any environmental observations, if applicable.

$OSA.agecomp

data.frame returned by TMB::oneStepPredict for age composition observations conditional on any aggregate catch or index, or environmental observations, if applicable.

$osa

One-step-ahead residuals (if do.osa=TRUE)

See Also

fit_wham

Examples

## Not run: 
data("input4_SNEMAYT") # load SNEMA yellowtail flounder data and parameter settings
mod <- fit_wham(input4_SNEMAYT, do.osa =FALSE, do.retro =FALSE)
mod <- make_osa_residuals(mod) # calculate Mohn's rho
plot_wham_output(mod)

## End(Not run)

timjmiller/wham documentation built on Nov. 23, 2023, 2:46 a.m.