The function attempts to find the optimal size for VCG.

Description

The function tries out different sizes of VCG and searches for the smallest distance.

Usage

BestVCGsize(formula, data = data, plot = TRUE)

Arguments

Details

It is only intended for exploratory purposes, as the VCG size is normally given. But it can be used to see how well the given size fits. The recommendation for VCG size is based solely on distance and does not take into account other aspects such as power or validity.

Value

If 'plot = TRUE', returns a list with:

Examples

set.seed(2342)
dat <- data.frame(
  treat = rep(0:1, c(50, 30)),
  cov1 = c(rnorm(50, 11, 2),  rnorm(30, 10, 1)),
  cov2 = c(rnorm(50, 12, 2),  rnorm(30, 10, 1)),
  cov3 = c(rnorm(50, 9,  2),  rnorm(30, 10, 1))
)
 BestVCGsize(treat ~ cov1 + cov2 + cov3, data=dat)

VCG Sampler for Energy Distance Balancing

Description

This function performs energy distance based balancing and selects a subset from pool based on energy distance to approximate a randomized control trial. Optionally, it visualizes the balancing results.

Usage

VCG_sampler(formula, data, n, c_w = NULL, random = FALSE, plot = TRUE)

Arguments

Details

If random is set to FALSE, the function selects the top 'n' units from the pool with the lowest energy distance and assigns them to the VCG group. If random is set to TRUE, the function samples 'n' units from pool with sampling probability inversely proportional to energy distance. The quality of covariate balancing is visualized using differences in medians and median absolute deviations (MADs). Permutation ellipses are generated by randomly permuting the pool and treated groups to estimate usual (random) variability. Only the X and Y axes are computed directly; the ellipse is interpolated between the axes. This method is intended as a visual approximation rather than a precise statistical test.

Value

If 'plot = TRUE', returns a list with:

• A data frame with added columns:

  • 'VCG': Indicator for selected pool units. VCG==1 indicates the VCG selected.

  • 'e_weights': Energy weights used for selection

  • '<treated>_balanced': A factor indicating balanced treated assignment.

• A ggplot2 object showing the median and MAD differences before and after balancing, with a 95

If 'plot = FALSE', returns only the modified data frame.

Examples

dat   <- data.frame(
  cov1  = rnorm(50, 10, 1),
  cov2  = rnorm(50, 7,  1),
  cov3  = rnorm(50, 5,  1),
  treated = rep(c(0, 1), c(35, 15))
)
  VCG_sampler(treated ~ cov1 + cov2 + cov3, data=dat, n=5)

Combine data from pool and treated groups

Description

If your data is stored in separate files, you can use this function to merge them.

Usage

combine_data(POOL_data, TG_data, indicator_name = "treated")

Arguments

Value

Data frame with all covariates that were present in both files and with new indicator factor POOL vs TG

Examples

pool_data <- data.frame(
  cov1   = rnorm(100, 11, 2),
  cov2   = rnorm(100, 11, 2),
  cov3   = rnorm(100, 11, 2),
  sex    = rbinom(100, 1, 0.5))

tg_data <- data.frame(
  cov2   = rnorm(20, 12, 1),
  cov3   = rnorm(20, 12, 1),
  cov4   = rnorm(20, 12, 1),
  sex    = rbinom(20, 1, 0.5))

 dx <- combine_data(pool_data, tg_data)
 str(dx)

Compute Energy Distance Between Two Groups

Description

Calculates the energy distance between two groups.

Usage

energy_distance(formula, data, standardized = TRUE)

Arguments

Details

Energy distance is a non-parametric measure of distributional difference. It is sensitive to differences in location, scale, and shape between groups. Before calculation, the covariates are scaled to a mean value of 0 and a standard deviation of 1.

Value

A numeric value representing the energy distance between the two groups.

Examples

dat <- data.frame(
 treated = rep(0:1, c(50, 30)),
 age    = c(rnorm(50, 5, 2),   rnorm(30, 5, 1)),
 weight = c(rnorm(50, 11, 2),  rnorm(30, 10, 1)),
 class  = c(rbinom(50, 3, 0.6),   rbinom(30, 3, 0.4))
 )

 energy_distance(treated ~ age + weight + class, data=dat)

Permutation Energy Test for Covariate Imbalance

Description

Performs a permutation-based energy distance test to assess whether two groups (defined by a binary treated variable) are balanced across a set of covariates. Optionally, it visualizes the distribution of permuted energy distances and highlights the observed test statistic and critical value.

Usage

energy_test(formula, data, alpha = 0.05, R = 2000, plot = TRUE)

Arguments

Details

The energy distance is a non-parametric measure of distributional difference. This test evaluates whether the covariate distributions between two groups are statistically distinguishable. A small p-value indicates imbalance between groups. A one-sided test is used because the energy distance is strictly positive; only values greater than the observed statistic in the permutation distribution are relevant.

Value

If 'plot = TRUE', returns a list with:

• A list of class '"htest"' containing:

  • 'p.value': The permutation p-value.

  • 'estimate': The observed energy distance.

  • 'critical.value': The critical value at the specified alpha level.

  • 'alternative': The alternative hypothesis ("one.sided").

  • 'method': Description of the test.

  • 'n.permutations': Number of permutations performed.

  • 'permutations': Vector of permuted energy distances.

• A ggplot2 object showing the histogram of permuted distances, with vertical lines for the observed statistic and critical value.

If 'plot = FALSE', returns only the '"htest"' result list.

See Also

element

Examples

dat <- data.frame(
 treated = rep(0:1, c(50, 30)),
 age    = c(rnorm(50, 5, 2),   rnorm(30, 5, 1)),
 weight = c(rnorm(50, 11, 2),  rnorm(30, 10, 1)),
 class  = c(rbinom(50, 3, 0.6),   rbinom(30, 3, 0.4))
 )

 energy_test(treated ~ age + weight + class, data=dat, R = 500)

Multi-Sample VCG Generator and Overlap Visualization

Description

Repeatedly samples VCGs (via 'VCG_sampler' and 'random=TRUE') from the pool, optionally plots the overlap of VCGs.

Usage

multiSampler(formula, data, n, c_w = NULL, Nsamples = 20, plot = TRUE)

Arguments

Details

The function repeatedly calls 'VCG_sampler' with 'random' set to TRUE to generate multiple VCG groups. It calculates the frequency of selection for each observation and computes the average percentage of overlapping observations. This function should only be used if you really need multiple VCG, e.g. for PoC studies. It is not intended for selecting one VCG from them afterwards! In this case, the VCG_sampler function should be used directly and only one VCG should be generated.

Value

If 'plot = TRUE', returns a list with:

data

The original data frame with additional VCG columns ('VCG_1', ..., 'VCG_Nsamples').

p

A 'ggplot2' object showing the number of times each observation was selected across VCG samples.

If 'plot = FALSE', returns the modified data frame only.

Examples

  dat <- data.frame(
  treat = rep(0:1, c(50, 30)),
  cov_1 = c(rnorm(50, 5, 2),   rnorm(30, 5, 1)),
  cov_2 = c(rnorm(50, 11, 2),  rnorm(30, 10, 1))
  )

  result <- multiSampler(treat ~ cov_1 + cov_2, data = dat, n = 10, Nsamples = 10)

Visualize Covariate Distribution Across TG, VCG, and POOL

Description

Creates a plot to compare the distribution of a selected variable across three groups: TG (treated groups), VCG (virtual control group), and POOL (data pool).

Usage

plot_var(data, what = NULL, stratum = "in_stratum", group = "VCG", title = "")

Arguments

Details

The function uses energy distance to quantify distributional differences between groups. For continuous variables, it overlays dashed lines for TG group statistics (mean, min, max) and displays sample sizes. For categorical variables, it uses color-coded bars and cumulative proportion lines to highlight imbalance.

Value

A ggplot2 object showing either:

• A boxplot for continuous variables (more than 4 unique values).

• A proportional bar chart for categorical variables (2–4 unique values).

Examples

dat   <- data.frame(
  cov1  = rnorm(50, 10, 1),
  cov2  = rnorm(50, 7,  1),
  cov3  = rnorm(50, 5,  1),
  treated = rep(c(0, 1), c(35, 15))
)
  out <- VCG_sampler(treated ~ cov1 + cov2 + cov3, data=dat, n=5, plot=FALSE)
  plot_var(out, what='cov1', group='VCG')
  plot_var(out, what='cov2', group='VCG')

Robust Scaling of Numeric and Categorical Variables

Description

Applies robust scaling to numeric and categorical variables. For numeric variables, the function centers by the median and scales by the MAD. For categorical variables with 2–4 unique levels, it applies a custom transformation to map them to numeric values.

Usage

robust_scale(x, group)

Arguments

Details

This function is designed to make numeric and categorical variables comparable. This is an internal function that should not be used by package users.

Value

A scaled numeric vector or a data frame with scaled columns.

Examples

dat<-data.frame(x=rnorm(100, 10, 3), sex=factor(rbinom(100, 1, 0.5), labels=c("M","F")))

x<- robust_scale(dat$x, dat$sex)
round(median(x), 2)
round(mad(x), 2)