Skip to contents

Changelog

Source: NEWS.md

tidystata (development version)

  • Shared help pages compare related functions and aliases in opening bullets. Grouping, labels, cutting edge cases and examples reflect current behavior.

  • isid() confirms successful key checks with a message while returning the original data invisibly. Use suppressMessages() for silent execution. .missing replaces .missok without an alias; update existing calls. Empty strings, including empty-string factor values, are rejected by default; .missing = TRUE allows them and other missing values but still requires unique key combinations.

  • capture() returns the expression’s result on success and invisible NULL on error, replacing its previous TRUE/FALSE result. It preserves successful visibility and evaluates once. Caught errors emit a failure message by default; quiet = TRUE suppresses it along with warnings and messages from the expression. Printed output is retained. Put assignments inside capture() to preserve existing objects on failure. Existing Boolean uses must be updated.

  • corr() calculates Pearson correlations on a common complete-case sample, like Stata’s corr (correlate), with tidy selection, computed columns, groups, weights, and long or wide tibble output.

tidystata 0.3.2

  • assert() protects input columns from internal-name collisions and rejects matrix conditions instead of reporting nonexistent row numbers.

  • summ() excludes zero-weight observations from weighted arithmetic, scales large analytic weights safely, and avoids overflow when averaging percentile boundaries. Weighted percentiles retain their values for large finite weights. Datetimes with an empty timezone attribute are accepted. Summary and tabulation help distinguish analytic weights from survey-design inference.

  • summ(), the tabulation family and pwcorr() retain a grouping column when it is also the weight column in an implicit all-column selection.

  • stata_merge() retains shared non-key values on using-only rows while preserving master values, including missing values, on matched rows.

  • cut_length() and cut_breaks() omit SPSS user-missing values before computing intervals. Binning functions reject integer64 inputs with an explicit-conversion message instead of reading their underlying payloads.

  • row_min() and row_max() recognize .data[[column]] inputs in one-row groups and keep .env scalars from filling otherwise all-missing rows.

  • tostring() writes numbers in full, as Stata does: 100000 gives "100000", not "1e+05". destring() reads labelled text like plain text, so ignore, its warning and readr::problems() apply, and rejects a doubled sign such as "-$-5".

  • misstable() counts SPSS user-missing codes and ranges. List columns use the established summary rule: scalar NA elements are missing and NULL elements are observed values. Grouping names that collide with its result columns give a clear error.

  • row_min()/row_max() treat .values columns as columns in one-row groups, and a missing constant makes rows missing when na.rm = FALSE, as in pmin().

  • stata_merge() keeps the master key when using has a non-key column of the same name; cut_breaks() warns when the default median does not split the values and rejects missing breaks and all-missing input.

  • Fixes: tab2() and fre2() no longer tabulate grouping or weight columns without .by, keep factor order around blank headers and compute Cramer’s V for large 2 x 2 tables; summ(.missing = FALSE) accepts weight vectors; as_dummy() handles columns without observed values; is_first() accepts pick(); print_all() accepts width; malformed varlist() ranges, NA positions and var_type(abbr = NA) give clear errors; pc() is missing for an infinite total; the timer no longer prints 60.00s.

tidystata 0.3.1

  • findit() corrects audited statistical mappings, missing-value handling and estimator defaults, with explicit bilingual qualifications and command-specific sources. Name substitution, documented abbreviations, row-number queries, by-prefix options and Chinese chart questions work while example anchors stay stable. Earlier curated examples show their notes in compact output too.

  • Add a searchable English/Chinese Stata-to-R reference book, generated from the findit() catalogue, with chapter navigation and a matching-page language switch. A dedicated pkgdown navigation link opens the book on the author’s website.

  • Expand the bilingual findit() reference with executable, source-linked examples and selected alternative R approaches. The generated guide provides an alphabetical command index, subject tables and a package-sorted R index.

  • Keep concise defaults while showing qualifications and sources on explicit expansion. Stable example IDs allow several R answers for the same Stata code.

  • Respect package-qualified R queries, avoid unrelated word-fragment matches, and explain known gaps and unrepresented option combinations.

  • findit() shows a single clear default, retaining alternatives for ambiguous inputs. Comma-separated variable names and hints customize catalogue examples, as in findit(tab, sex, region) and findit(tab, gen); all matches remain available through print(n = Inf).

tidystata 0.3.0

  • Renamed from statart to tidystata, with a focus on data exploration and preparation. Legacy import/export, regression and plotting wrappers are removed.
  • findit() translates Stata commands, R function names and plain questions into working examples, with English and Chinese guides. In consoles with colour, the Stata and R code retain automatic console colours. Results start with compact example blocks, omit redundant function notes and count remaining examples explicitly.
  • Expanded tabulation and summaries include cross-tables, weights, correlations, missing-value reports and grouped calculations.
  • Added key checks, duplicate handling, Stata-style merges, panel operators, variable transformations and persistent table previews.
  • Statistical conventions follow Stata and are checked against its output. Tables preserve the input container; missing values and labels are handled consistently. See the articles and function reference for details.
  • Faster row calculations and grouped aggregation, with fewer required dependencies. Requires R 4.1.0 and dplyr 1.1.4 or later; flextable is optional.

tidystata 0.2.0

  • Initial CRAN submission.