baseline_equivalence() gains missing-data handling options (#6): a
missing argument choosing between "pairwise" (the previous behavior and
still the default: each covariate uses its own available cases) and
"complete" (every row restricted to the cases complete on the treatment
column and all evaluated covariates, so every row describes the same set
of cases), and a report_missing argument adding missing_treatment /
missing_comparison columns with each group's share of missing values,
computed before any deletion. gt_baseline() labels and formats the new
columns. Default output is unchanged.wwc_robustness(): reports how stable a baseline-equivalence verdict is to
the computation choices an analyst could defensibly make differently
(standardizing by the pooled versus the comparison-group standard deviation,
and applying the WWC small-sample correction or not), flagging whether each
covariate's category, and the overall verdict, changes. A multiverse view of a
single WWC determination.attrition_boundary(): classifies a study as low or high attrition against
the WWC attrition boundary (Standards Handbook v4.1, Table II.1), under the
cautious or optimistic assumption. Complements attrition(), which reports the
rates but leaves the classification to the user.wwc_rating(): applies the WWC group-design rating logic (Standards
Handbook v4.1, Section II) to attrition and baseline equivalence, returning
"Meets Without Reservations", "Meets With Reservations", or "Does Not Meet".cluster_correction(): the WWC clustering correction for mismatched
analyses (Procedures Handbook v4.1, Appendix F, after Hedges 2007). Corrects
the t statistic and its degrees of freedom for clustering and returns the
clustering-corrected p value and significance decision, with the WWC default
ICCs (0.20 achievement, 0.10 other). Validated against 1,800+ clustered
findings in the official WWC study database.cox_index() follows Procedures Handbook v4.1 [VI.1.2] exactly:
the Cox index carries no small-sample correction (that correction applies to
Hedges' g only).hedges_g() and cox_index() now handle missing values consistently when
na.rm = FALSE: both reject missing input up front with the message
"Missing values present; set na.rm = TRUE to drop them." Previously
hedges_g() raised an opaque internal error while cox_index() silently
returned NA (#12).wwc_summary(): collapse a baseline_equivalence() table into an overall
WWC verdict (satisfied / satisfied_with_adjustment / not_satisfied) plus
per-category counts and the largest absolute effect size.attrition(): overall and differential attrition for a two-group design,
the inputs to the WWC attrition standard.tutoring: a simulated quasi-experimental tutoring
evaluation (400 students) whose covariates span all three WWC equivalence
categories. See data-raw/tutoring.R for how it is generated.love_plot(): a Love plot of standardized effect sizes across covariates,
with WWC threshold reference lines and points coloured by category (requires
ggplot2).gt_baseline(): render a baseline equivalence table as a formatted gt
table with readable labels and rounded statistics (requires gt).ggplot2 and gt added to Suggests; both functions error gracefully if the
package is not installed.cox_index(): WWC Cox index standardized effect size for binary
(dichotomous) covariates.baseline_equivalence() now handles binary covariates (numeric 0/1,
logical, or two-level factor) via the Cox index, in addition to continuous
covariates via Hedges' g. A covariate with exactly two unique values is
treated as binary.type column ("continuous" / "binary"), and the
effect-size column is renamed from hedges_g to the estimator-agnostic
effect_size (breaking change). For binary covariates, mean_treatment
and mean_comparison report event proportions.gt / flextable).Initial scaffold.
hedges_g(): standardized mean difference (Hedges' g) between a treatment
and a comparison group, with the WWC small-sample correction factor.wwc_classify(): classify standardized mean differences into the three WWC
baseline-equivalence categories.baseline_equivalence(): build a report-ready baseline equivalence table for
continuous covariates.