NEWS

palsr 0.1.1

New rolling_origin_pals(): a genuine out-of-sample (rolling-origin) evaluation, re-estimating the model at each of a set of origin dates from strictly earlier events and scoring predictions on the following holdout window. Complements the one-step-ahead ("marching forward") criterion that estimate_pals() minimizes, in which every event's own location contributes to the single fitted parameter vector.
bootstrap_pals() gains a block argument for a block bootstrap (block = "year", or a user-supplied grouping vector) that preserves within-block temporal/actor dependence, as an alternative to the default i.i.d. event bootstrap.
README: fixed the pool_rubin() quick-start example, which errored with "unused arguments" when called with q =/u = (the function's arguments are named estimates/variances); updated install instructions to lead with CRAN and to add build_vignettes = TRUE to the GitHub install example, so vignette("palsr") is available after following the README.

Initial release implementing the Projected Actor Location (PALS) method of Kim, Liu and Desmarais (2023) doi:10.1017/psrm.2022.6.
Core workflow: pal_events(), estimate_pals() (one- and four-parameter models), project_pal() / project_pals(), predict_event_locations(), and pal_distance().
Uncertainty quantification via bootstrap_pals() and Rubin's Rules pooling with pool_rubin().
Fast great-circle distance and projection kernels in C++ via Rcpp.
Bundled simulated example dataset nigeria_sim and simulate_conflict_events().
Introductory vignette and full function documentation.