Changes in version 1.1.3                        

  - New vignette on (semi-)supervised learning
  - Fix parameter counts for models with constraints
  - Use safe Hessian inversion even for models without random effects in
    HMM$post_coeff() and HMM$confint()

                 Changes in version 1.1.2 (2025-11-05)                  

  - Add hurdle negative binomial distribution
  - Fix bug in HMM$fit_stan()
  - Fix bug in HMM$pseudores()

                 Changes in version 1.1.1 (2025-09-11)                  

  - Add error message for laplace = TRUE in HMM$fit_stan()
  - Add reference to JSS paper

                 Changes in version 1.1.0 (2025-06-24)                  

  - Add standard error to output of HMM$confint()
  - Use ID-specific delta0 when initial_state = "stationary"
  - Add par_alt() functions for pretty display of cat and mvn
    parameters.
  - Improve mvn distribution (automatically detect dimension), and fix
    mvnorm for dim > 2, using a Cholesky decomposition to get
    unconstrained parameters of covariance matrix.
  - Add zero-one-inflated beta distribution
  - Replace optimx by nlminb for model fitting
  - Allow empty models for simulation
  - Use generalized determinant for penalty matrices
  - Add nbinom(mean, shape) distribution
  - Allow for user-specified arguments to be passed directly to
    mgcv::gam(); e.g., knots.
  - Add built-in functions dvm, rvm, dwrpcauchy, rwrpcauchy, to remove
    dependence on CircStats (as requested by CRAN).

                 Changes in version 1.0.2 (2023-10-24)                  

  - Fix bug for categorical distribution
  - Fix bug when using tibble input data
  - Fix bugs for models with shared parameters
  - Add HMM$suggest_initial()