Planned for the next release (0.3.0):
analyze_target() ranks every feature by its association with a chosen target
column, adapting the measure to the variable types (absolute Pearson
correlation, the correlation ratio eta, or Cramer's V) so numeric and
categorical predictors are comparable on a common 0-1 scale. plot_target()
visualises the ranking.compare_datasets() detects distribution drift between two data frames
(e.g. train vs test), reporting the population stability index (PSI) plus a
Kolmogorov-Smirnov (numeric) or chi-squared (categorical) test per column.
plot_drift() charts the per-column PSI.Changes requested during the initial CRAN review:
normality_tests() no longer touches the global random-number state. Large
columns are now reduced with a deterministic, evenly-spaced subsample instead
of set.seed() + sample(); the seed argument has been removed.New analysis and reporting:
report() renders a complete profile to a self-contained HTML file
(requires pandoc, via \pkg{rmarkdown}).categorical_association() and plot_association() add Cramer's V between
categorical columns (the categorical analogue of the correlation matrix).analyze_dates() profiles date/datetime columns: range, unique count, and
the largest gap between consecutive timestamps.compare_groups() summarises numeric columns within the levels of a grouping
column (grouped/comparative profiling).Pipeline changes:
profile_data() gains group_by (adds a grouped comparison to the
diagnostics) and distributions (set FALSE to skip the eager per-column
distribution plots on wide data). Association and date results are now part of
the returned object, and plot() accepts which = "association".summary() now also prints date, association and grouped-comparison sections
when present.profile_data() with type inference, missing-value analysis,
summary statistics (incl. skewness/kurtosis), normality tests, outlier
detection (IQR/z-score/robust), correlation analysis, a data-quality score,
and ggplot2 figures, returned as a data_profile S3 object with print(),
summary() and plot() methods.