statsExpressions 2.2.1

• one_sample_test() with a one-sided alternative now also reports a
  one-sided confidence interval for the parametric and nonparametric effect
  sizes. Previously the p-value was one-sided but the effect size
  confidence interval was always two-sided, unlike in two_sample_test().

• contingency_table() no longer fails for a test of independence when x
  or y has an empty level, either an unused factor level or one emptied by
  removing rows with missing values. Such levels are now dropped, as in
  pairwise_contingency_table().

• For between-subjects designs (paired = FALSE), long_to_wide_converter()
  now drops a level of x whose responses are all missing. Previously the
  empty level was kept, which made robust oneway_anova() and
  pairwise_comparisons() fail.

• tidy_model_expressions() no longer duplicates rows and mismatches
  expressions when term values repeat, as in multivariate models where each
  term appears once per response.