| Title: | Multivariate Outlier Detection Methods |
|---|---|
| Description: | Provides tools for detecting multivariate outliers in numeric datasets using Mahalanobis distance, robust Minimum Covariance Determinant (MCD), and Principal Component Analysis (PCA)-based methods. The Mahalanobis distance calculations are performed using an efficient C++ backend via Rcpp. |
| Authors: | Senuri Yasara [aut, cre] |
| Maintainer: | Senuri Yasara <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.0.0.9000 |
| Built: | 2026-06-16 09:49:00 UTC |
| Source: | https://github.com/senuyasara/multivariate_outlier_detection_r_package |
Detects multivariate outliers using Mahalanobis, Minimum Covariance Determinant (MCD), or PCA-based distances. Supports robust detection by computing distance scores for each observation and comparing them against a chi-squared cutoff at a specified significance level.
detect_multivariate_outliers(data, method = "mahalanobis", alpha = 0.975)detect_multivariate_outliers(data, method = "mahalanobis", alpha = 0.975)
data |
A numeric data frame or matrix. |
method |
Outlier detection method: "mahalanobis", "mcd", or "pca". |
alpha |
Significance level (default = 0.975). |
A data frame combining the original input data with distances and outlier flags.
df_mtcars <- mtcars[, c("mpg", "hp", "wt" )] head(df_mtcars) ## Mahalanobis Distance result_mahal <- detect_multivariate_outliers(df_mtcars, method = "mahalanobis", alpha = 0.975) ## Minimum Covariance Determinant (MCD) result_mcd <- detect_multivariate_outliers(df_mtcars, method = "mcd", alpha = 0.975) ## Principal Component Analysis (PCA) result_pca <- detect_multivariate_outliers(df_mtcars, method = "pca", alpha = 0.975)df_mtcars <- mtcars[, c("mpg", "hp", "wt" )] head(df_mtcars) ## Mahalanobis Distance result_mahal <- detect_multivariate_outliers(df_mtcars, method = "mahalanobis", alpha = 0.975) ## Minimum Covariance Determinant (MCD) result_mcd <- detect_multivariate_outliers(df_mtcars, method = "mcd", alpha = 0.975) ## Principal Component Analysis (PCA) result_pca <- detect_multivariate_outliers(df_mtcars, method = "pca", alpha = 0.975)
Generates 2D scatterplots for each pair of variables in the dataset, with outliers identified using Mahalanobis or MCD distances computed across all variables, without including each observation in its own distance calculation.
plot_outliers(data, method = c("mahalanobis", "mcd"), alpha = 0.975)plot_outliers(data, method = c("mahalanobis", "mcd"), alpha = 0.975)
data |
A numeric data frame or matrix. |
method |
Outlier detection method: "mahalanobis" or "mcd". |
alpha |
The quantile cutoff for identifying outliers (default 0.975). |
df_mtcars <- mtcars[, c("mpg", "hp", "wt" )] head(df_mtcars) ## Pairwise Plots: Mahalanobis plot_outliers(df_mtcars, method = "mahalanobis", alpha = 0.975) ## Pairwise Plots: MCD plot_outliers(df_mtcars, method = "mcd", alpha = 0.975)df_mtcars <- mtcars[, c("mpg", "hp", "wt" )] head(df_mtcars) ## Pairwise Plots: Mahalanobis plot_outliers(df_mtcars, method = "mahalanobis", alpha = 0.975) ## Pairwise Plots: MCD plot_outliers(df_mtcars, method = "mcd", alpha = 0.975)