Plot objects

Overview

The lionfish package offers multiple different types of display elements. These have to be provided to the interactive_tour function that launches the GUI. Each plot object is a list containing a type and an object. The type specifies, which kind of display should be generated. The object provides additional information required to construct the display. The plot objects then have to be stored in a list that can be given to the interactive_tour function. The currently supported displays are:

• 1d tour
• 2d tour
• scatterplot
• histogram
• mosaic plot
• heatmap
• categorical cluster interface

Examples for all displays can be found below.

Setup

# Load required libraries
library(tourr)
library(reticulate)
library(lionfish)
library(flexclust)

# Initialize python backend
init_env()

Dataset 1 - Flea data

# Load dataset
data("flea")

# Prepare objects for later us
clusters_flea <- as.numeric(flea$species)
flea_subspecies <- unique(flea$species)

# Standardize data and calculate half_range
flea <- apply(flea[,1:6], 2, function(x) (x-mean(x))/sd(x))
feature_names_flea <- colnames(flea)
half_range_flea <- max(sqrt(rowSums(flea^2)))

Dataset 2 - Winter activities data

# Load dataset and set seed
data("winterActiv")
set.seed(42)

# Perform kmeans clustering 
clusters_winter <- stepcclust(winterActiv, k=6, nrep=20)
clusters_winter <- clusters_winter@cluster

# Get the names of our features
feature_names_winter <- colnames(winterActiv)

Currently supported display types

One dimensional tour

To display one dimensional tours, they first have to be generated and saved using the save_history function of the tourr package. For more information on the available types of tours please visit tour path construction.

Example plot object

guided_tour_flea_1d <- save_history(flea,
                                    tour_path=guided_tour(holes(),1))

obj_flea_1d_tour <- list(type="1d_tour",          # type of display
                         obj=guided_tour_flea_1d) # 1d tour history 

Interactivity

• Changeing frames of tour: The frames of the current tour can be changed by pressing the left arrow-key for the last frame and the right arrow-key for the next frame. The current frame is displayed in the frame on the left.
• Projection manipulation: the projection can be manipulated clicking on one of the arrowheads with the right mouse-button and dragging it.
• Subselection: The datapoins within each subset can be changed by clicking on one of the bars of the histogram. The datapoints within the bar are then moved to the subset currently selected on the left.

1d tour with flea dataset

Two dimensional tour

To display two dimensional tours, they first have to be generated and saved using the save_history function of the tourr package. For more information on the available types of tours please visit tour path construction.

Example plot object

grand_tour_flea_2d <- save_history(flea,
                                   tour_path = grand_tour(d=2))

obj_flea_2d_tour <- list(type="2d_tour",         # type of display
                         obj=grand_tour_flea_2d) # 2d tour history

Interactivity

• Changeing frames of tour: The frames of the current tour can be changed by pressing the left arrow-key for the last frame and the right arrow-key for the next frame. The current frame is displayed in the frame on the left.
• Projection manipulation: the projection can be manipulated clicking on one of the arrowheads with the right mouse-button and dragging it.
• Subselection: The datapoins within each subset can be changed by pressing the left mouse-button and drawing a lasso around the datapoints to be selected. The datapoints within the lasso are then moved to the subset currently selected on the left.

2d tour with flea dataset

Scatterplot

To display a scatter plot, the features to be displayed on the x and y axis have to be provided in form of a two dimensional vector.

Example plot object

obj_flea_scatter <- list(type="scatter",          # type of display
                         obj=c("tars1", "tars2")) # x and y axis of plot

Interactivity

• Subselection: The datapoins within each subset can be changed by pressing the left mouse-button and drawing a lasso around the datapoints to be selected. The datapoints within the lasso are then moved to the subset currently selected on the left.

Scatterplot with tars1 and tars2 of flea dataset

Histogram

To display a histogram, the feature to be displayed on the x axis has to be provided in form of a two dimensional vector.

Example plot object

obj_flea_histogram <- list(type="hist", # type of display
                           obj="head")  # x axis of histogram

Interactivity

• Subselection: The datapoins within each subset can be changed by clicking on one of the bars of the histogram. The datapoints within the bar are then moved to the subset currently selected on the left.

Histogram with head of flea dataset

Mosaic

To display a mosaic plot, one has to provide whether the subgroups/clusters should be on the x or y axis, either with “subgroups_on_y” or “subgroups_on_x”.

Example plot object

obj_winter_mosaic <- list(type="mosaic",   # type of display
                  obj=c("subgroups_on_y")) # whether subgroups x or y axis

Interactivity

Currently there is no way of directly interacting with the mosaic plot.

Moisaic plot of winter activities dataset

Heatmap

To display a heatmap, the metric to be calculated and plotted has to be selected. One can choose between “Total fraction”, “Intra cluster fraction” and “Intra feature fraction”.

Consider the matrix $$C = \left[ \begin{array}{cccc} c_{11} & c_{12} & \dots & c_{1p} \\ c_{21} & c_{22} & \dots & c_{2p} \\ \vdots & \vdots & \ddots & \vdots \\ c_{k1} & c_{k2} & \dots & c_{kp} \end{array} \right]$$

where $c_{ij}, i=1, ..., k$ (number of clusters); $j=1, ..., p$ (number of features) are a summary of each feature in each cluster. In case of binary data $c_{ij}$ are the positive counts of the cluster/feature combination.

Then the total fraction is calculated by $f_{ij}^{o} = \frac{c_{ij}}{n}$

the intra cluster fraction by $f_{ij}^{c} = \frac{c_{ij}}{n_{i}}$ and

the intra feature fraction by $f_{ij}^{f} = \frac{c_{ij}}{n_{j}}$,

where $n_i, n_j$ are the row and column totals.

Example plot object

obj_winter_heatmap <- list(type="heatmap",          # type of display
                           obj=c("Total fraction")) # initial metric

Interactivity

• Metric selection: The currently displayed metric can be changed by using the dropdown menu within the frame on the left. The details on the metrics can be found above.

Heatmap of winter activities dataset

Categorical cluster interface

To display a categorical cluster interface, the metric to be calculated and plotted has to be selected. One can choose between “Total fraction”, “Intra cluster fraction” and “Intra feature fraction”. For details see “Heatmap”

Example plot object

obj_winter_cat_clust <- list(type="cat_clust_interface", # type of display
                             obj=c("Total fraction"))    # initial metric

Interactivity

• Metric selection: The currently displayed metric can be changed by using the dropdown menu within the frame on the left. The details on the metrics can be found above.

Categorical cluster interface of winter activities dataset

Generating the displays

The various plot objects can the be displayed with the interactive_tour function.

# interactive_tour call of flea dataset. Insert plot objects of your liking.
interactive_tour(data=flea,
                 feature_names=feature_names_flea,
                 plot_objects=list(obj_flea_2d_tour),
                 half_range=half_range_flea,
                 n_plot_cols=2,
                 preselection=clusters_flea,
                 n_subsets=3,
                 preselection_names=flea_subspecies,
                 display_size=5)

# interactive_tour call of winterActiv dataset. Insert plot objects of your liking.
interactive_tour(data=winterActiv,
                 feature_names=feature_names_winter,
                 plot_objects=list(obj_winter_cat_clust),
                 half_range=3,
                 n_plot_cols=2,
                 preselection=clusters_winter,
                 n_subsets=10,
                 display_size=5)