Evaluate tag observations under a continuous-time latent growth process

Propagate interpolated release-length states through a continuous-time generator, then integrate a Normal or Student-t observation density over the latent final-length states. The Student-t standard deviation is observation error only; process variation is represented by the generator.

Usage

get_growth_tag_likelihood_ctmc(
  generator,
  midpoint,
  initial,
  time_index,
  time_step,
  release_length,
  observed_final_length,
  observation_sd,
  observation_model = 3L,
  df = 3,
  tolerance = 1e-12
)

Arguments

generator: A column-oriented continuous-time growth generator.
midpoint: Length-state representative values (mm).
initial: Matrix of release-state probabilities, one tag per column.
time_index: Positive integer time-grid index for each tag.
time_step: Width of one time-grid step in years.
release_length: Observed release lengths (mm).
observed_final_length: Observed recapture lengths (mm).
observation_sd: Actual observation-error SD (mm).
observation_model: 0 for no tag likelihood, 1 for Normal, or 3 for Student-t.
df: Student-t degrees of freedom; ignored for other models.
tolerance: Numerical tolerance passed to RTMB::expAv().

Value

A list containing expected increments, latent process SDs, and tag log likelihoods.