klct v0.2.0

Statistical method improvements

• Welch's t-test by default: All parametric pairwise comparisons now
  always use Welch's t-test (unequal variances assumed). The preliminary
  F-test for equal variances (var.test()) has been removed, as it inflates
  Type I error rates. This affects compare_groups(), compare_groups_2x2(),
  and plot_groups().

• Type III sums of squares: compare_groups_2x2() now uses
  lm() + car::Anova(type = "III") with sum-to-zero contrasts instead of
  stats::aov() (which uses Type I sequential SS). This gives correct tests
  for main effects and interactions with unbalanced designs.

• Benjamini--Hochberg adjusted p-values: A P.adj column is now
  included in all results tables. In compare_groups(), pairwise p-values
  are BH-adjusted per sex across all measures at a site. In
  compare_groups_2x2(), omnibus p-values are BH-adjusted per factor across
  measures, and pairwise p-values are BH-adjusted per sex across measures.
  Significance stars (Sig) and red-color highlighting in the R Markdown
  templates continue to be based on the raw (unadjusted) p-value, as the
  report is intended for exploratory screening.

New output columns

• Difference column: A Diff column shows the raw difference (group 2
  minus group 1). For parametric results this is the difference in means; for
  nonparametric results, the Hodges--Lehmann pseudomedian of pairwise
  differences (from wilcox.test(conf.int = TRUE)).

• 95% confidence intervals: CI.Low and CI.High columns report the
  95% CI bounds for the difference. For parametric comparisons these come from
  Welch's t-test. For nonparametric comparisons these come from the Wilcoxon
  rank-sum test with conf.int = TRUE.

• Effect size reporting: A Pct.Change column shows the percent change
  of the second group relative to the first. For parametric results this is
  based on group means; for nonparametric results, on group medians.

• SD column: Parametric results now include an SD column alongside the
  existing SEM column.

• Q1 and Q3 columns: Nonparametric results now report Q1 (25th
  percentile) and Q3 (75th percentile) instead of IQR, giving a clearer
  picture of the distribution's location and spread.

Dependencies

• Added car to Imports (for car::Anova() with Type III sums of squares).

Template changes

• The omnibus test label in print_results_2x2() now reads "Type III ANOVA"
  (parametric) and the pairwise label reads "Welch's t-test" to reflect the
  updated methods.
• R Markdown templates no longer require all three data files (spine, metaphysis,
  diaphysis) to be present. When a data file is missing, the corresponding
  section is omitted from the report entirely rather than causing the report to
  error. Only the mouse table is required.
• The knit: function reference in standard (non-Mayo) templates has been
  corrected from microctr::knit_with_colored_text to
  klct::knit_with_colored_text.