Package {vizClust}

Interactive plot of bootstrap confidence intervals

Description

Generates an interactive plot displaying the statistic value and its 95% bootstrap confidence intervals at each step of the transition between two clusters.

Usage

get_interval(data, nn_idx, B = 1000, vars = NULL, n_vars = NULL)

Arguments

Details

The function performs the following steps:

1. Extracts the k-nearest neighbors for each step along the transition.

2. Computes bootstrap samples of the mean for each variable.

3. Estimates 95% confidence intervals using the bootstrap distribution.

4. Generates an interactive plot displaying the mean trajectories together with their confidence intervals.

Value

An interactive visualization displaying the trajectories of the selected variables across transition steps, together with bootstrap-based 95% confidence intervals and interactive tooltips containing variable names and interval values.

See Also

pseudosamples(), knn_statistics(), plot_explorer()

Examples

## Load example dataset
data(iris)

## Keep only numeric variables and scale
iris_scaled <- as.data.frame(scale(iris[, -5]))

## Perform PAM clustering
set.seed(123)
pam_iris <- cluster::pam(iris_scaled, k = 2)

## Extract medoids and generate pseudo-samples
medoids <- pam_iris$medoids
pseudo  <- pseudosamples(medoids, c1 = 1, c2 = 2, n_points = 20)

## Run KNN statistics
knn_res <- knn_statistics(iris_scaled, pseudo, k = 5, fun = "mean")

## Plot with bootstrap confidence intervals for all variables
get_interval(iris_scaled, knn_res$nn_idx, B = 100)

## Plot top 2 variables by variance
get_interval(iris_scaled, knn_res$nn_idx, B = 100,
             n_vars = 2)  ## Results for top variance variables

## Plot specific variables
get_interval(iris_scaled, knn_res$nn_idx, B = 100,
             vars = c("Sepal.Length", "Sepal.Width"))  ## Results for selected variables

KNN-based summary statistics for pseudo-samples

Description

This function maps pseudo-samples onto real data using k-nearest neighbors (KNN) and computes summary statistics for each variable, including the mean, median and standard deviation.

Usage

knn_statistics(data, pseudo.sample, k, fun = "mean")

Arguments

Details

For each pseudo-sample, the function identifies the k nearest neighbors in the original dataset and computes a summary statistic for each variable across the selected neighbors.

Supported summary statistics include:

• "mean": mean of the neighboring observations.

• "median": median of the neighboring observations.

• "sd": standard deviation of the neighboring observations.

Value

A list containing:

See Also

pseudosamples(), get_interval(), plot_explorer()

Examples

## Load example dataset
data(iris)

## Keep only numeric variables and scale
iris_scaled <- as.data.frame(scale(iris[, -5]))

## Perform PAM clustering
set.seed(123)
pam_iris <- cluster::pam(iris_scaled, k = 2)

## Extract medoids and generate pseudo-samples
medoids <- pam_iris$medoids
pseudo  <- pseudosamples(medoids, c1 = 1, c2 = 2, n_points = 20)

## Run KNN statistics with mean summary
knn_res <- knn_statistics(iris_scaled, pseudo, k = 15, fun = "mean")

head(knn_res$explorer)  ## Results for the explorer data frame
head(knn_res$nn_idx)    ## Results for the KNN indices

Visualization of variable trajectories across cluster transitions

Description

This function visualizes the evolution of selected variables along the pseudo-sample path using an interactive line plot.

Usage

plot_explorer(explorer, vars = NULL, n_vars = NULL)

Arguments

Details

This function generates an interactive line plot showing how selected variables evolve along the transition path defined by the pseudo-samples.

The input explorer is typically obtained from knn_statistics(), where rows represent transition steps between clusters and columns represent variables.

Variable selection can be controlled as follows:

• If vars is provided, only the specified variables are displayed.

• If n_vars is provided, the variables with the highest variance across the transition are selected.

• If neither argument is provided, all variables are displayed.

The function reshapes the data into long format and creates an interactive visualization using ggplot2 and ggiraph, allowing users to explore variable trajectories dynamically.

Value

An interactive ggiraph object representing a line plot of variable trajectories across the transition.

Each line corresponds to a variable, and each point along the x-axis represents a transition step between clusters.

See Also

pseudosamples(), knn_statistics(), get_interval()

Examples

## Load example dataset
data(iris)

## Keep only numeric variables and scale
iris_scaled <- as.data.frame(scale(iris[, -5]))

## Perform PAM clustering
set.seed(123)
pam_iris <- cluster::pam(iris_scaled, k = 2)

## Extract medoids and generate pseudo-samples
medoids <- pam_iris$medoids
pseudo  <- pseudosamples(medoids, c1 = 1, c2 = 2, n_points = 20)

## Run KNN statistics
knn_res <- knn_statistics(iris_scaled, pseudo, k = 15, fun = "mean")

## Plot all variables
plot_explorer(knn_res$explorer)

## Plot specific variables
plot_explorer(knn_res$explorer,
              vars = c("Sepal.Length", "Sepal.Width"))  ## Results for selected variables

## Plot top 2 variables by variance
plot_explorer(knn_res$explorer,
              n_vars = 2)  ## Results for top variance variables

Pseudo-sample generation between cluster medoids

Description

This function generates interpolated pseudo-samples along the linear transition between two cluster medoids. It is useful for exploring transitions between clusters in a multivariate feature space.

Usage

pseudosamples(medoids, c1, c2, n_points)

Arguments

Details

The function computes a linear interpolation between two cluster medoids. A sequence of values for lambda between 0 and 1 is generated, and for each value, a new pseudo-sample is calculated as:

medoid\_c1 + \lambda * (medoid\_c2 - medoid\_c1)

This procedure produces a continuous trajectory in the feature space between the two clusters.

Value

A data frame with n_points rows and the same number of columns as medoids. Each row represents a pseudo-sample along the transition path between the two clusters.

See Also

knn_statistics(), plot_explorer(), get_interval()

Examples

## Load example dataset
data(iris)

## Keep only numeric variables and scale
iris_scaled <- scale(iris[, -5])

## Perform PAM clustering
set.seed(123)
pam_iris <- cluster::pam(iris_scaled, k = 2)

## Extract medoids
medoids <- pam_iris$medoids

## Generate pseudo-samples between cluster 1 and 2
pseudo <- pseudosamples(medoids, c1 = 1, c2 = 2, n_points = 20)
head(pseudo)