Skip to contents

Changelog

Source: NEWS.md

rankingQ 0.2.0

  • First CRAN submission. The package was previously available on GitHub under version 0.1.0.
  • The bundled identity and identity_w datasets now use item names for the marginal-rank columns: party, religion, gender, race (previously app_identity_1–app_identity_4) and household, neighborhood, city, state (previously anc_identity_1–anc_identity_4). Examples and vignettes now pass these columns explicitly, e.g. main_q = c("party", "religion", "gender", "race").
  • New exported function rank_wider() for converting long-format ranking data to wide format (one column per item or a single pasted ranking string).
  • imprr_direct(): anc_correct is now optional. Users can instead supply p_random directly, or rely on anchor-based estimation of the non-random response rate.
  • New exported convenience wrapper add_ipw_weights() for returning the original data augmented with respondent-level IPW weights, with an option to also keep ranking summaries.
  • imprr_direct_rcpp(): correctly delegates to imprr_direct() when method = "fixed", forwarding all population and assumption inputs.
  • rpluce(): when all remaining Plackett-Luce weights become zero after earlier draws, the remaining items are now sampled uniformly at random rather than being ordered by input position.
  • rank_longer(): stricter validation on the reference argument and clearer handling of single- vs multi-column input.
  • Stronger input validation across the package: weights must be finite, non-negative, and sum to a positive number; J must be finite; permutation labels are parsed consistently across compact, delimited, and fixed-width formats (including J > 9).
  • Permutation-label infrastructure rewritten to support rankings of arbitrary size.
  • Documentation: switched to \href{}{} for external URLs; expanded examples and \value{} sections.

rankingQ 0.1.0