Rita: Automated Transformations, Normality Testing, and Reporting
Automatically evaluates common normalizing transformations of numeric
variables and selects the best-fitting transformation for each based
on squared correlation with theoretical normal quantiles. Available
transformations include Rankit, arcsine, logit/log-odds, logarithmic,
square-root, and inverse/reciprocal. Performs six normality tests
(Thode, 2002, <doi:10.1201/9780203910894>), incorporating
simulation-based p-value calibration for small-sample Jarque–Bera and
optionally Lilliefors, as well as subsampling for Shapiro–Wilk when
sample sizes exceed its supported limit. Available tests include
Anderson–Darling, Pearson chi-square, D'Agostino–Pearson omnibus,
Jarque–Bera, Kolmogorov–Smirnov with Lilliefors correction, and
Shapiro–Wilk. A descriptive statistical report of the retained input
data in their original units, including the five-number summary and
mathematical moments, is also presented. The Rankit empirical normal
quantile transformation (ENQT) (Soloman and Sawilowsky, 2009,
<doi:10.22237/jmasm/1257034080>) is provided for those requiring
rank-based normalization and to facilitate its adoption. Structured
analytical results and paired visualizations of original and transformed
distributions are also available.
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=Rita
to link to this page.