silk

The goal of silk is to make curves with outliers smooth as silk.

Installation

You can install the development version of silk from GitHub with:

# install.packages("devtools")
devtools::install_github("TeroJii/silk")

Usage

The objective of this package is to provide a simple way to correct individual outlier values. The package includes two main functions: find_outliers() and correct_outliers(), for finding and correcting outliers, respectively.

Let’s take a look at these functions in action.

Detecting outliers in example data

The package contains an example dataset silk_data1 which we can use to demonstrate the functionality of the package.

library(silk)
## basic example code
data("silk_data1")

silk_data1 |> 
  head()
#>   time         y   group
#> 1    1  8.584244 series1
#> 2    2  9.159694 series1
#> 3    3  9.717704 series1
#> 4    4 10.249923 series1
#> 5    5 10.748432 series1
#> 6    6 11.205878 series1

If we visualize the data, we see that there are a few unusual values.

library(ggplot2)

silk_data1 |> 
  ggplot2::ggplot(ggplot2::aes(x = time, y = y)) +
  ggplot2::geom_line() +
  ggplot2::theme_bw() +
  ggplot2::facet_wrap(~group)

We can detect these unusual data points using the silk::find_outliers() function, which takes a data.frame and a column name as input, and returns a modified version of the input data.frame.

silk_data1 |> 
  silk::find_outliers(col = "y", group_col = "group") |>
  head()
#>   time         y   group  .median     .mad .outlier
#> 1    1  8.584244 series1 8.065949 2.713192    FALSE
#> 2    2  9.159694 series1 8.065949 2.713192    FALSE
#> 3    3  9.717704 series1 8.065949 2.713192    FALSE
#> 4    4 10.249923 series1 8.065949 2.713192    FALSE
#> 5    5 10.748432 series1 8.065949 2.713192    FALSE
#> 6    6 11.205878 series1 8.065949 2.713192    FALSE

As we can see, the function has added new columns to the data. The .outlier column indicates whether a value is an outlier or not.

Visualizing the outliers

We can visualize the outliers using the ggplot2 package:

silk_data1 |> 
  silk::find_outliers(col = "y", group_col = "group") |>
  ggplot2::ggplot(ggplot2::aes(x = time, y = y)) +
  ggplot2::geom_line() +
  ggplot2::geom_point(ggplot2::aes(color = .outlier)) +
  ggplot2::theme_bw() +
  ggplot2::facet_wrap(~group) +
  ggplot2::theme(legend.position = "top")

Correcting the outliers

The silk package also includes a function for correcting the outliers. The silk::correct_outliers() function can be used together with the silk::find_outliers() function in the following fashion:

silk_data1 |> 
  silk::find_outliers(col = "y", group_col = "group") |> 
  silk::correct_outliers(col = "y") |> 
  ggplot2::ggplot(ggplot2::aes(x = time, y = y)) +
  ggplot2::geom_line() +
  ggplot2::theme_bw() +
  ggplot2::facet_wrap(~group) +
  ggplot2::ggtitle("Corrected data")

The silk::correct_outliers() function overwrites the original column values, replacing the outlier values with the mean of the previous and following observation in the data.