Skip to contents

Changelog

Source: NEWS.md

bboutools 1.3.0

Breaking changes

  • Models now use CaribouYear instead of Year throughout (#75). This affects column names in output data frames from predict(), augment(), and all bb_predict_*() functions. Existing bboutools code (e.g., bb_fit_survival(), bb_predict_survival()) is unaffected, but custom code that references Year columns in output data frames will need updating.
  • Multi-population models change parameter indexing from bAnnual[1] to bAnnual[1,1] and add a PopulationName column to output data frames (#77). Single-population models are also affected. As above, this does not break standard bboutools workflows but may affect custom code that extracts parameters or manipulates output data frames directly.

New features

  • bb_fit_survival() and bb_fit_recruitment() now support multiple populations in a single model fit (#77). Population-level intercepts (b0[k]) and year random effects (bAnnual[i,k]) are estimated per population, while month effects and variance components are shared. bb_predict_growth() and bb_predict_population_change() auto-filter to shared population and CaribouYear combinations when survival and recruitment fits differ.
  • bb_fit_survival() now accepts aggregate annual survival data (one row per population per year) when monthly collar data is unavailable (#78). Annual data is auto-detected. The model skips the month random effect and does not apply monthly-to-annual exponentiation.
  • bb_fit_survival() and bb_fit_recruitment() now support prediction of unobserved years via allow_missing = TRUE (#79). Users add placeholder rows with PopulationName and Year filled and all measurement columns NA_integer_. Predictions for unobserved years draw the year random effect from its Normal(0, sAnnual) prior, giving appropriate uncertainty. An informational message lists detected unobserved CaribouYears by population when quiet = FALSE.
  • Added bb_priors_survival_national() and bb_priors_recruitment_national() to generate disturbance-informed intercept priors using the bbouNationalPriors package (#80).
  • Added bb_predict_*_samples() variants for all bb_predict_*() functions to return raw MCMC samples instead of summary tables (#83).
  • bb_fit_survival() and bb_fit_recruitment() now accept niters = 0 to build the model without MCMC sampling, useful for inspecting model structure (#85).
  • Added prior-only sampling for prior predictive checks (#91). Users provide all-placeholder data with allow_missing = TRUE and niters > 0 to sample from priors without observed data.
  • Plot functions now automatically facet by PopulationName when multiple populations are present.

Deprecated

Bug fixes

  • Fixed ML model fitting error with nimble >= 1.4.0 by using nimbleQuad for Laplace approximation (#88).
  • Added informative error when nimble is not attached to the search path, e.g., when using bboutools:: instead of library(bboutools) (#70).

Internal

  • model_code() now shows resolved numeric prior values instead of symbolic names (e.g., dnorm(3, sd = 10) instead of dnorm(b0_mu, sd = b0_sd)) and removes redundant { } blocks left by NIMBLE’s if/else resolution (#69).
  • Migrated user-facing messages from message()/warning()/abort_chk() to cli package (#90). .inform_unobserved_years() and .warn_filtered_multi() use plain message() for clean display in bboushiny toast notifications.
  • Reduced sysdata.rda size from ~7.7 MB to ~890 KB.

Documentation

  • Updated Get Started, Analytical Methods, and Priors vignettes for multi-population models, annual data, and national priors (#82).
  • Added Extensions article covering multi-population analysis, aggregate annual survival, unobserved year predictions, prior-only sampling, and raw MCMC samples (#82).

bboutools 1.2.0

bboutools 1.1.0

bboutools 1.0.0

  • cleaned up documentation following JOSS review.
  • add Stable lifecycle badge.

bboutools 0.1.0

  • added rhat argument to glance() and converged() to adjust convergence threshold, with default value of 1.05. Convergence no longer depends on ess.
  • removed exclude_year argument in bb_fit_survival_ml() as failure to fit fixed year effect resolved with intercept adjustment.
  • provide warning to user if start of caribou year (year_start) not matching for survival and recruitment model fits in bb_predict_growth().

bboutools 0.0.0.9007

bboutools 0.0.0.9006

  • Moved internal data checking functions to be exported function in bboudata.
  • Updated bbtutorial vignette to mention the new data checking functions.

bboutools 0.0.0.9005

  • bb_fit_recruitment() will error if there are any NA values in the PopulationName, Year, Month and Day columns.
  • bb_fit_survival() will error if there are any NA values in the PopulationName, Year, Month, StartTotal, MortalitiesCertain and MortalitiesUncertain columns.

bboutools 0.0.0.9004

Added ‘Getting Started’ vignette.

Added functions:

Added arguments:

bboutools 0.0.0.9003

  • Switch MCMC engine from Jags to Nimble.
  • Added recruitment/survival Maximum Likelihood fit functions:
  • Added summary methods for Maximum Likelihood output with class ‘bboufit_ml’.
  • Changed how priors are passed to fit functions (e.g. priors = c(b0_mu = 4, b0_sd = 2) instead of priors = list(b0 = "dnorm(4, 2^-2)")).
  • Added argument sig_fig to coef()/tidy()/predict() methods to round estimates and lower/upper bounds to set significant figures.

bboutools 0.0.0.9002

  • bb_predict_growth() now handles different years and doesn’t return month.
  • Adjusted default priors.

bboutools 0.0.0.9001

bboutools 0.0.0.9000

  • Added a NEWS.md file to track changes to the package.