                        Changes in version 0.1.0                        

Initial CRAN release.

  - Model fitting and prediction for exposure-response models of
    time-to-event endpoints: ertte_aft() (a parametric accelerated
    failure time model built on survival::survreg(), supporting
    exponential, Weibull, log-normal, and log-logistic distributions)
    and ertte_coxph() (a semi-parametric proportional-hazards model
    built on survival::coxph()).
  - ertte_aft_select_distribution(), AIC-based selection across the
    supported AFT base distributions.
  - ertte_predict() and ertte_fun() for survival-probability prediction
    and counterfactual curve evaluation without refitting, for both
    engines.
  - ertte_landmark() and ertte_rmst(), reducing a fitted survival curve
    to a binary landmark response or a restricted mean survival time,
    with confidence intervals.
  - simulate.ertte_model(), drawing simulated event/censoring times from
    a fitted model, with an optional per-row administrative censor_time.
  - Stepwise covariate modelling (ertte_scm_forward(),
    ertte_scm_backward(), ertte_scm_history()), built on the single-term
    ertte_add_term()/ertte_remove_term() helpers, with a choice of
    "p-value", "aic", or "bic" selection criterion.
  - ertte_power(), a power-function covariate transform usable as a
    model term in either engine's formula or an SCM candidate set.
  - Interoperability with the companion erplots package via
    er_predict()/er_simulate()/er_summary()/ er_predict_survival()
    methods, registered lazily so ertte has no hard dependency on
    erplots or on any plotting package.
  - An example dataset, ertte_data, simulated from a Weibull AFT ground
    truth with an exposure and sex effect, administrative censoring, and
    independent dropout censoring.
