What’s new in chisquare
version 1.2
Statistical methodology
improvements
- Exact p-value calculation for permutation and Monte Carlo tests:
Implemented the Phipson & Smyth (2010) method for computing exact
p-values when permutations are randomly drawn. The formula p = (B +
1)/(m + 1) ensures that p-values are never zero and that Type I error
rates are correctly controlled. This supersedes the previous
implementation that could underestimate p-values and inflate Type I
error rates;
New association
measures and effect sizes
- Maximum-corrected version of the C coefficient added;
- 95 percent confidence interval around Goodman-Kruskal lambda
(asymmetric and symmetric) added;
- 95 percent confidence interval around Goodman-Kruskal tau
added;
- Enhanced effect size interpretation for chi-square-based measures:
uncorrected measures (Phi, C, Cramer’s V) are now assessed against
Cohen’s thresholds adjusted by their maximum achievable values, while
corrected measures use standard Cohen’s thresholds;
- Added a detailed effect size interpretation thresholds table showing
exact threshold values for all association measures;
New residual analysis
methods
- Goodman-Kruskal residuals added;
- PEM (Percentage of Maximum Deviation from Independence) added;
- Standardised median polish residuals and adjusted standardised
median polish residuals added, providing robust alternatives for cell
residual analysis that are resistant to masking and swamping effects
from outliers;
User interface and
output improvements
- Added a new
gt.tables component to the returned results
that contains all formatted gt tables; this allows the tables to be
rendered any time without needing to re-run the analysis;
- Modification to the code to avoid automatic rendering of all tables
in the R console, improving execution speed and preventing browser
overload; tables are now returned in a structured list (see previous
bullet point) but only rendered when setting the new
render.all.tables parameter to TRUE;
Documentation improvements
- Major restructuring of the help documentation for improved clarity
and pedagogical flow: content is now organised into six thematic
sections (Hypothesis Testing and Power Analysis; Decomposition of the
Chi-Square Statistic; Table Standardisation; Chi-Square-Maximising
Table; Post-Hoc Analysis; Measures of Categorical Association);
- Table standardisation methodology (Iterative Proportional Fitting)
now documented separately from specific coefficient applications,
enhancing conceptual clarity;
- Association measures reorganised by statistical foundation:
Chi-square-based measures, Margin-free measures, and PRE (Proportional
Reduction in Error) measures, facilitating coefficient selection based
on analytical needs;
Bug fixes and minor
improvements
- Annotation at the bottom of the last rendered table removed since no
longer relevant;
- Typo in the title of the absolute contribution to the chi-square
statistic table fixed;
- Updates to the help documentation’s content and layout;
- Bibliographical references updated.