confidenceellipse

The ConfidenceEllipse package provides functions for computing the coordinate points of confidence ellipses and ellipsoids for a given bivariate and trivariate dataset, at user-defined confidence level.

https://github.com/christiangoueguel/confidenceellipse

Keywords

Repository

The ConfidenceEllipse package provides functions for computing the coordinate points of confidence ellipses and ellipsoids for a given bivariate and trivariate dataset, at user-defined confidence level.

README.Rmd

---
output: github_document
---



```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%",
  fig.retina = 2
)
```




# 

[![R-CMD-check](https://github.com/ChristianGoueguel/ConfidenceEllipse/actions/workflows/R-CMD-check.yaml/badge.svg)](https://github.com/ChristianGoueguel/ConfidenceEllipse/actions/workflows/R-CMD-check.yaml)
[![Project Status: Active – The project has reached a stable, usable
state and is being actively
developed.](https://www.repostatus.org/badges/latest/active.svg)](https://www.repostatus.org/#active)
[![Lifecycle:
stable](https://img.shields.io/badge/lifecycle-stable-brightgreen.svg)](https://lifecycle.r-lib.org/articles/stages.html#stable)
[![License:
MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
[![codecov](https://codecov.io/gh/ChristianGoueguel/ConfidenceEllipse/graph/badge.svg?token=JAGMXN2F70)](https://app.codecov.io/gh/ChristianGoueguel/ConfidenceEllipse)
[![](https://cranlogs.r-pkg.org/badges/last-week/ConfidenceEllipse?color=orange)](https://cran.r-project.org/package=ConfidenceEllipse)
[![](https://cranlogs.r-pkg.org/badges/ConfidenceEllipse?color=yellowgreen)](https://cran.r-project.org/package=ConfidenceEllipse)
[![](https://cranlogs.r-pkg.org/badges/grand-total/ConfidenceEllipse)](https://cran.r-project.org/package=ConfidenceEllipse)
[![CRAN
status](https://www.r-pkg.org/badges/version/ConfidenceEllipse)](https://CRAN.R-project.org/package=ConfidenceEllipse)


The `ConfidenceEllipse` package computes the coordinate points of confidence region for a given bivariate and trivariate dataset. The size of the elliptical region is determined by the confidence level, and the shape is determined by the covariance matrix. The confidence level, typically 95% or 99%, indicates the probability that the ellipse contains the true mean vector, assuming a distribution.

## Installation

You can install `ConfidenceEllipse` from CRAN using:

```r
install.packages("ConfidenceEllipse")
```

Alternatively you can grab the development version from github using devtools:

```r
install.packages("devtools")
devtools::install_github("ChristianGoueguel/ConfidenceEllipse")
```

## Example

```{r message=FALSE, warning=FALSE}
library(magrittr)
library(tidyselect)
library(patchwork)
library(dplyr)
library(ggplot2)
library(purrr)
library(tidyr)
```

```{r message=FALSE, warning=FALSE}
library(ConfidenceEllipse)
```

### Dataset

The dataset is comprised of 13 different measurements for 180
archaeological glass vessels from different groups (Janssen, K.H.A., De
Raedt, I., Schalm, O., Veeckman, J.: Microchim. Acta 15 (suppl.) (1998)
253-267. Compositions of 15th - 17th century archaeological glass
vessels excavated in Antwerp.)

```{r}
data("glass", package = "ConfidenceEllipse")
```

```{r}
glass %>% glimpse()
```

### Confidence Region
#### Classical and robust confidence ellipses

First, the `confidence_ellipse` function is used to compute coordinate points of the confidence ellipse and then the ellipse is plotted on a two-dimensional plot `x`-`y` of the data. It shows a region that, with a specified confidence level `conf_level`, is expected to contain the population mean vector of a bivariate distribution.

```{r message=FALSE, warning=FALSE}
ellipse_95 <- confidence_ellipse(glass, x = SiO2, y = Na2O, conf_level = 0.95, robust = FALSE, distribution = "normal")
rob_ellipse_95 <- confidence_ellipse(glass, x = SiO2, y = Na2O, conf_level = 0.95, robust = TRUE, distribution = "normal")
```

```{r message=FALSE, warning=FALSE}
ellipse_95 %>% glimpse()
```

```{r}
cutoff <- qchisq(0.95, df = 2)
MDsquared <- glass %>%
  select(SiO2, Na2O) %>%
  as.matrix() %>%
  mahalanobis(colMeans(.), cov(.), inverted = FALSE)
```

```{r}
plot1 <-
  ggplot() +
  geom_path(data = ellipse_95, aes(x = x, y = y), color = "blue", linewidth = 1L) +
  geom_point(data = glass %>% mutate(md = MDsquared) %>% filter(md <= cutoff), aes(x = SiO2, y = Na2O), shape = 21L, color = "black", fill = "lightblue", size = 3L) +
  geom_point(data = glass %>% mutate(md = MDsquared) %>% filter(md > cutoff), aes(x = SiO2, y = Na2O), shape = 21L, color = "black", fill = "gold", size = 3L) +
  labs(x = "SiO2 (wt.%)", y = "Na2O (wt.%)", title = "Classical confidence ellipse\nat 95% confidence level") +
  theme_bw() +
  theme(
    panel.grid = element_blank(),
    legend.position = "none"
  )
```

```{r}
x_mcd <- glass %>%
  select(SiO2, Na2O) %>%
  as.matrix() %>%
  robustbase::covMcd()
```

```{r}
rob_MDsquared <- glass %>%
  select(SiO2, Na2O) %>%
  as.matrix() %>%
  mahalanobis(x_mcd$center, x_mcd$cov)
```

```{r}
plot2 <-
  ggplot() +
  geom_path(data = rob_ellipse_95, aes(x = x, y = y), color = "blue", linewidth = 1L) +
  geom_point(data = glass %>% mutate(md = rob_MDsquared) %>% filter(md <= cutoff), aes(x = SiO2, y = Na2O), shape = 21L, color = "black", fill = "lightblue", size = 3L) +
  geom_point(data = glass %>% mutate(md = rob_MDsquared) %>% filter(md > cutoff), aes(x = SiO2, y = Na2O), shape = 21L, color = "black", fill = "gold", size = 3L) +
  labs(x = "SiO2 (wt.%)", y = "Na2O (wt.%)", title = "Robust confidence ellipse\nat 95% confidence level") +
  theme_bw() +
  theme(
    panel.grid = element_blank(),
    legend.position = "none"
  )
```

```{r}
plot1 | plot2
```

##### Grouping

For grouping bivariate data, the `.group_by` argument can be used if the data contains an unique grouping variable (`.group_by = NULL` by default). When a grouping variable is provided, the function will compute the ellipses separately for each level of the factor, allowing you to explore potential differences or patterns within subgroups of the data.

It's important to note that the grouping variable should be appropriately coded as a factor before passing it to the `.group_by` argument. If the variable is currently stored as a character or numeric type, you may need to convert it to a factor using functions like `as.factor()` or `forcats::as_factor()`.

```{r}
rpca_scores <- glass %>%
  select(where(is.numeric)) %>%
  pcaPP::PCAproj(method = "qn") %>%
  pluck("scores") %>%
  as_tibble() %>%
  mutate(glassType = glass %>% pull(glassType)) %>%
  rename(PC1 = Comp.1, PC2 = Comp.2)
```

```{r message=FALSE, warning=FALSE}
ellipse_pca <- rpca_scores %>% confidence_ellipse(x = PC1, y = PC2, .group_by = glassType)
```

```{r}
ggplot() +
  geom_point(data = rpca_scores, aes(x = PC1, y = PC2, color = glassType, shape = glassType), size = 3L) +
  geom_path(data = ellipse_pca, aes(x = x, y = y, color = glassType), linewidth = 1L) +
  scale_color_brewer(palette = "Set1", direction = 1) +
  labs(x = "PC1", y = "PC2", title = "Principal component analysis") +
  theme_bw() +
  theme(
    aspect.ratio = .7,
    panel.grid = element_blank(),
    legend.position = "none"
  )
```

#### Ellipsoid

The `confidence_ellipsoid` function accepts an additional variable `z` and computes the ellipsoid for trivariate data.

```{r message=FALSE, warning=FALSE}
ellipsoid_grp <- glass %>% confidence_ellipsoid(SiO2, Na2O, Fe2O3, glassType)
```

```{r}
ellipsoid_grp %>% glimpse()
```

```{r message=FALSE, warning=FALSE}
rgl::setupKnitr(autoprint = TRUE)
rgl::plot3d(
  x = ellipsoid_grp$x,
  y = ellipsoid_grp$y,
  z = ellipsoid_grp$z,
  xlab = "SiO2 (wt.%)",
  ylab = "Na2O (wt.%)",
  zlab = "Fe2O3 (wt.%)",
  type = "s",
  radius = 0.03,
  col = as.numeric(ellipsoid_grp$glassType)
)
rgl::points3d(
  x = glass$SiO2,
  y = glass$Na2O,
  z = glass$Fe2O3,
  col = as.numeric(glass$glassType),
  size = 5
)
rgl::view3d(theta = 260, phi = 30, fov = 60, zoom = .85)
```

Owner

GitHub Events

Packages