Internal function to compute OSA residuals for a single composition slice

Backend function used by [get_osa()] to calculate one-step-ahead (OSA) residuals from observed and expected composition data.

Usage

run_osa(
  obs,
  exp,
  N = NULL,
  DM_theta = NULL,
  LN_Sigma = NULL,
  fleet,
  index,
  years,
  index_label,
  comp_like
)

Arguments

obs

Matrix of observed compositions (rows = years, columns = bins)

exp

Matrix of expected compositions (same shape as obs)

N

Sample size for multinomial/Dirichlet-multinomial

DM_theta

Dirichlet-multinomial overdispersion parameter(s)

LN_Sigma

Logistic-normal covariance matrix

fleet

Fleet identifier

index

Vector of composition bins (last entry dropped for logistic-normal)

years

Vector of years corresponding to rows of obs/exp

index_label

Character describing ages or lengths

comp_like

Integer specifying likelihood:

0

multinomial

1

Dirichlet-multinomial

2-4

logistic-normal variants

Value

A list containing:

res

Data frame of OSA residuals with columns fleet, index_label, year, index, resid