Skip to contents

Calculating effect sizes

Source: R/001_effsize_func.R
effect_size.Rd

Computes the effect size for each control-test group pairing in idx. The resampling bootstrap distribution of the effect size is then subjected to Bias-corrected and accelerated bootstrap (BCa) correction.

The following effect sizes mean_diff, median_diff, cohens_d, hedges_g and cliffs_delta are used for most plot types.

Usage

mean_diff(dabest_obj, perm_count = 5000)

median_diff(dabest_obj, perm_count = 5000)

cohens_d(dabest_obj, perm_count = 5000)

hedges_g(dabest_obj, perm_count = 5000)

cliffs_delta(dabest_obj, perm_count = 5000)

cohens_h(dabest_obj, perm_count = 5000)

Arguments

dabest_obj

A dabest_obj created by loading in dataset along with other specified parameters with the load() function.

perm_count

The number of reshuffles of control and test labels to be performed for each p-value.

Value

Returns a dabest_effectsize_obj list with 22 elements. The following are the elements contained within:

  • raw_data The tidy dataset passed to load() that was cleaned and altered for plotting.

  • idx The list of control-test groupings as initially passed to load().

  • delta_x_labels Vector containing labels for the x-axis of the delta plot.

  • delta_y_labels String label for the y-axis of the delta plot.

  • Ns List of labels for x-axis of the raw plot.

  • raw_y_labels Vector containing labels for the y-axis of the raw plot.

  • is_paired Boolean value determining if it is a paired plot.

  • is_colour Boolean value determining if there is a colour column for the plot.

  • paired Paired ("sequential" or "baseline") as initially passed to load().

  • resamples The number of resamples to be used to generate the effect size bootstraps.

  • control_summary Numeric value for plotting of control summary lines for float_contrast = TRUE.

  • test_summary Numeric value for plotting of control summary lines for float_contrast = TRUE.

  • ylim Vector containing the y limits for the raw plot.

  • enquo_x Quosure of x as initially passed to load().

  • enquo_y Quosure of y as initially passed to load().

  • enquo_id_col Quosure of id_col as initially passed to load().

  • enquo_colour Quosure of colour as initially passed to load().

  • proportional Boolean value as initially passed to load().

  • minimeta Boolean value as initially passed to load().

  • delta Boolean value as initially passed to load().

  • proportional_data List of calculations related to the plotting of proportion plots.

  • boot_result List containing values related to the calculation of the effect sizes, bootstrapping and BCa correction.

  • baseline_ec_boot_result List containing values related to the calculation of the effect sizes, bootstrapping and BCa correction for the baseline error curve.

  • permtest_pvals List containing values related to the calculations of permutation t tests and the corresponding p values, and p values for different types of effect sizes and different statistical tests.

Details

The plot types listed under here are limited to use only the following effect sizes.

  • Proportion plots offers only mean_diff and cohens_h.

  • Mini-Meta Delta plots offers only mean_diff.

The other plots are able to use all given basic effect sizes as listed in the Description.

Examples

# Loading of the dataset
data(non_proportional_data)

# Applying effect size to the dabest object
dabest_obj <- load(non_proportional_data,
  x = Group, y = Measurement,
  idx = c("Control 1", "Test 1")
)
dabest_obj.mean_diff <- mean_diff(dabest_obj)

# Printing dabest effectsize object
print(dabest_obj.mean_diff)
#> DABESTR v2023.9.12
#> ==================
#> 
#> Good morning!
#> The current time is 06:03 AM on Tuesday December 12, 2023.
#> 
#> The unpaired mean difference between Test 1 and Control 1 is 0.585 [95%CI 0.307, 0.869].
#> The p-value of the two-sided permutation t-test is 0.0022, calculated for legacy purposes only.
#> 
#> 5000 bootstrap samples were taken; the confidence interval is bias-corrected and accelerated.
#> Any p-value reported is the probability of observing the effect size (or greater),
#> assuming the null hypothesis of zero difference is true.
#> For each p-value, 5000 reshuffles of the control and test labels were performed.
#>