Run multiple t-tests by groups on tidy data

A wrapper around t_test_all_tidy that performs t-tests separately for each combination of grouping variables in a tidy dataset. Optionally, it provides Holm's step-down family-wise error control by returning per-hypothesis critical alpha thresholds.

Usage

t_test_all_tidy_grouped(
  dataset,
  n_group_col = 1,
  paired = F,
  var.equal = FALSE,
  onesample = FALSE,
  mu = 0,
  ci = c("freq", "bayes_central", "bayes_hdi"),
  alternative = c("two.sided", "less", "greater"),
  conf.level = 0.95,
  alpha = 0.05,
  holm = FALSE,
  pd = FALSE,
  bf = FALSE,
  cor = TRUE,
  mean_x_EAP = FALSE,
  mean_x_MAP = FALSE,
  mean_x_MED = FALSE,
  diff_EAP = FALSE,
  diff_MAP = FALSE,
  diff_MED = FALSE,
  cohens_d = NULL,
  cohens_d_EAP = FALSE,
  cohens_d_MAP = FALSE,
  cohens_d_MED = FALSE,
  cohens_dz = TRUE,
  cohens_dz_EAP = FALSE,
  cohens_dz_MAP = FALSE,
  cohens_dz_MED = FALSE,
  rscale_est = Inf,
  rscale_bf = "medium",
  iterations = 10000,
  map_density_n = 512,
  show_design = TRUE,
  detailed = FALSE,
  fullbayes = FALSE
)

Arguments

dataset: A tidy data frame containing participant ID, one or more grouping variables, an independent variable with exactly two levels, and a numeric dependent variable (in that order).
n_group_col: Integer. The number of grouping columns (default = 1).
paired: Logical. If TRUE, a paired t-test is performed (default FALSE).
var.equal: Logical. If TRUE, the two-sample t-test assumes equal variances (default FALSE).
onesample: Logical. If TRUE, a one-sample t-test is performed (default FALSE).
mu: Numeric. Null hypothesis value for the mean difference (default 0).
ci: Character vector specifying interval type(s): "freq", "bayes_central", or "bayes_hdi". Default is c("freq","bayes_central","bayes_hdi"). Passed to t_test_all.
alternative: Character. Alternative hypothesis: "two.sided", "less", or "greater". Default is "two.sided". Passed to t_test_all.
conf.level: Numeric. Confidence/credibility level (default 0.95). Passed to t_test_all.
alpha: Numeric in (0, 1). Family-wise significance level used for Holm's method when holm = TRUE. Default is 0.05.
holm: Logical. If TRUE and the output contains a p column, an alpha column is (re)computed using Holm's method via p_to_holmalpha with sig.level = alpha. This affects only the per-hypothesis critical alpha thresholds, not the p-values. Default is FALSE.
pd, bf, cor: Logical flags. Whether to compute probability of direction (pd), Bayes factors, or correlation (for paired samples). Passed to t_test_all.
mean_x_EAP, mean_x_MAP, mean_x_MED: Logical. Report posterior summaries for group means. Passed to t_test_all.
diff_EAP, diff_MAP, diff_MED: Logical. Report posterior summaries for the mean difference. Passed to t_test_all.
cohens_d: Logical or character. Type/request for Cohen's d in independent samples. Passed to t_test_all.
cohens_d_EAP, cohens_d_MAP, cohens_d_MED: Logical. Posterior summaries for Cohen's d. Passed to t_test_all.
cohens_dz: Logical. Request Cohen's dz for paired samples. Passed to t_test_all.
cohens_dz_EAP, cohens_dz_MAP, cohens_dz_MED: Logical. Posterior summaries for Cohen's dz. Passed to t_test_all.
rscale_est, rscale_bf: Numeric or character. Cauchy prior scales for estimation and Bayes factors (e.g., "ultrawide", "wide", "medium", or a positive number). Defaults are Inf and "medium", respectively. Passed to t_test_all.
iterations, map_density_n: Integer. MCMC iterations and grid size for MAP density. Passed to t_test_all.
show_design: Logical. If TRUE, show message of design (default TRUE).
detailed: Logical. Whether to return detailed results (TRUE) or minimal output (FALSE, default).
fullbayes: Logical. Whether to show only Bayesian results (TRUE) or both frequentist and Bayesian results (FALSE, default).

Value

A data frame with t-test results for each group, matching the structure of t_test_all_tidy() and, if requested, including Holm-based critical alpha thresholds in alpha.

Details

The function validates the dataset layout, groups by the specified grouping variables, and applies t_test_all_tidy() within each group. When holm = TRUE, it replaces/creates an alpha column containing Holm step-down critical alpha values corresponding to the reported p values (per group), controlling the family-wise error rate at alpha. P-values themselves are not adjusted.

Examples

# Example with one grouping variable:
head(data_fict_mixed)
#> # A tibble: 6 × 4
#>   participant gender condition       rt
#>         <dbl> <chr>  <chr>        <dbl>
#> 1           1 woman  control        679
#> 2           1 woman  experimental   900
#> 3           2 woman  control        712
#> 4           2 woman  experimental  1047
#> 5           3 woman  control        781
#> 6           3 woman  experimental  1056
t_test_all_tidy_grouped(data_fict_mixed, paired = TRUE, n_group_col = 1)
#> design: paired
#>   gender    diff      t df     p alpha sig cohens_dz
#> 1    man  -34.00  -0.43  3 0.699  0.05  ns    -0.213
#> 2  woman -277.25 -11.91  3 0.001  0.05   *    -5.954

Usage

Arguments

Value

Details

See also

Examples