eff_size: Effect size for difference between EDA objects.

In PiotrTymoszuk/ExDA: Toolset for Explorative Data Analysis

Effect size for difference between EDA objects.

Description

compare two or more EDA objects with a range of statistical tests.

Usage

eff_size(
  ...,
  type = c("cohen_d", "paired_cohen_d", "wilcoxon_r", "paired_wilcoxon_r", "cohen_kappa",
    "freeman_theta", "cramer_v", "cohen_kappa", "etasq", "partial_etasq",
    "kruskal_etasq", "kendall_w"),
  exact = TRUE,
  ci = TRUE,
  boot_method = c("percentile", "bca", "normality")
)

Arguments

...	EDA objects, at least two, created by eda.
type	type of statistic the effect size. For two numeric EDA objects: independent variable Cohen's D ('cohens_d', cohens_d), paired Cohen's D ('paired_cohen_d', cohen_d), independent variable Wilcoxon r ('wilcoxon_r', wilcox_effsize) or paired Wilcoxon r ('paired_wilcoxon_r', wilcox_effsize) are currently implemented. For two factor EDA objects: Cohen's Kappa ('cohen_kappa', Kappa) and Freeman's theta ('freeman_theta', freemanTheta) is implemented. For 2 or more factor EDAs: Cramer's V ('cramer_v', CramerV) can be calculated. For two or more numeric EDAs: ANOVA-derived (partial) eta-squared ('etasq' or 'partial_etasq', eta_squared) or Kruskal-Wallis-derived eta-squared ('kruskal_etasq', kruskal_effsize) and for aired data Kendall's W ('kendall_w', friedman_effsize) are currently available.
exact	logical, should exact values for Chi-squared. Mann-Whitney and Wilcoxon test be returned?
ci	logical, should confidence intervals for the test effect size be returned?
boot_method	indicates how the bootstrap confidence intervals are calculated. Can be any of 'percentile', 'bca', or 'normality', defaults to 'percentile'.

Details

EDA object type is coerced to factor or numeric, as appropriate for the requested analysis. The default statistic test results are returned as well: T test for Cohen's D, Mann-Whitney/Wilcoxon test for Wilcoxon r, Chi-squared test results for Cohen's kappa, Cramer's V and Freeman's theta, one-way ANOVA for eta-squared and partial eta-squared, Kruskal-Wallis test for Kruskal eta-squared and Friedman test for Kendall#s W. The p value referst to the statistical test result.

Value

an eTest object with the effect size statistic name as 'estimate' variable, effect size statistic value and with 95/

PiotrTymoszuk/ExDA documentation built on Nov. 17, 2024, 5:46 p.m.