Panning

To pan the plot, select the plot interior with the right (or secondary) mouse button and move the mouse (with the button still down). The direction of panning can be constrained by holding down the named modifier keys while panning.

Try it. Select anywhere on the scatterplot and move it

• freely about
• only horizontally
• only vertically

Question: How did the inspector change in response to each movement?

Reproduce these movements on the smaller “World View” plot appearing at the top of the inspector.

How does the larger scatterplot change?

Zooming

Another way to dynamically focus on different parts of the plot is by zooming, using scrolling:

To try it, note that there is a small cluster of 16 points towards the top left of the scatterplot.

On the scatterplot,

• zoom in and out and move the plot around until only these 16 points appear in the scatterplot

• use the restricted horizontal and vertical zooming so that the same 16 points occupy as much of the scatterplot’s space as possible

Question: How has the inspector changed in response?

On the inspector click on the button titled plot. What happened?

Using zooming and panning on the “World View”

• focus the scatterplot on the same 16 points.

Selecting

There are two ways to directly select points on the scatterplot using the mouse: either by selection (one at a time or multiple selections) or by “sweep selection” as shown below:

To try this out, on the scatterplot (which should contain only 16 points)

• place the mouse over the topmost point and click the left (or primary) mouse button
  • the point will be highlighted (magenta coloured)
• hold the shift key down and select the leftmost, rightmost and bottom most points of the 16
  • there should now be 4 magenta points
• colour these 4 points light blue and change their glyph to a triangle
  • click on the background of the plot to see the change in colour

Question: How has the inspector changed?

Now try a holding the left mouse button down and move the mouse.

This is called sweeping or sweep selection and allows us to sweep out a contiguous area of the plot.

• use sweep select to identify, and give different colours (none black) to 4 different (roughly) contiguous groups of your choice from the 16 points.

Check: The 16 points should now be separated by colour into 4 groups and by glyph shape into 2 groups.

Question: How has the inspector changed?

You will have already discovered how to colour selected points from the inspector. You can now use the inspector to select points by colour:

• from the inspector, select all the points of each colour in turn.

on brushing

You should have a scatterplot of open circles with three clusters of points in red, blue, and orange.

Use the functionality of the loon inspector to (in order)

1. Using only colour selection from the inspector, select the largest group (and no others)
2. Turn on brushing (as opposed to sweeping) mode
3. Under dynamic selection, choose deselect
4. Over the scatterplot use this brush mode to deselect the region nearest the highest density (shaped a bit like a question mark – ?)
5. Colour the selected points green (i.e. low density region of largest coloured group)

You should now have four different coloured groups or clusters.

For fun, in the inspector

1. select the green group
2. under dynamic selection, choose invert
3. move the brush over the scatterplot
4. move the brush over the scatterplot while holding the <Shift> key

Question: What is going on? When might this functionality be useful?

on selection

When done, turn the selection back to the defaults:

That is,

• [static:] none
• [dynamic:] select
• [by:] sweeping

Shift selection works with any mix of

• individual point selection
• static selection
• dynamic selection by sweeping or brushing
• by color

Exercise: Try out different combinations of the above selections.

moving points

For a variety of reasons, we sometimes want to (temporarily) move points in a scatterplot.

From the inspector:

1. deactivate all but the small red group
2. select the red group and scale the plot to the selected
3. with these selected points, explore the effect of each of the move modifications on the configuration of the selected points
4. describe in words what each one does

From the scatterplot, selected points may be moved by hand

• one at a time using <Ctrl>-<Left-button>
• in groups using <Shift><Ctrl><Left-button>

For example:

1. select the red group
2. move the red group to the bottom of the plot
3. return the red group to its original location

adding colours to the inspector

More colours can be added directly by clicking on the + or the +5 in the colour modification section of the inspector.

• clicking +5 twice adds 10 new colours
• these colours will persist with the inspector for every loon scatterplot the inspector palette of colours

The loon system default list of colours get be found programmatically:

head(l_getColorList())

[1] "gray60" "#A6CEE3" "pink" "#B2DF8A" "#FDBF6F" "#FFFF99"

Note:

• loon colours are 6 hexadecimal digits in length and can be a named R colour

• being constrained to tcl/tk colours, no alpha level is available at this time in loon

• R has a similar set of base colours that is default for all its graphics devices:

  head(palette())

  [1] "black" "#DF536B" "#61D04F" "#2297E6" "#28E2E5" "#CD0BBC"

• like the R palette, the loon inspector palette persists.

changing the inspector palette of colours

At the beginning of a loon session (that is, immediately after executing library(loon)) you might choose to change the default palette for the inspector to something that suits your tastes/problem.

Note: There may be unintended side effects to changing the inspector colours after a plot (and hence the inspector) has been created so this should be avoided.

A variety of functions exist to change an inspector’s palette of colours:

l_setColorList_baseR()  # base R palette
l_setColorList_ColorBrewer("Set2")  # colorblind friendly choice from ColorBrewer
l_setColorList_hcl(luminance = 80)  # set of hcl colours
l_setColorList_ggplot2()  # ggplot2's palette 
l_setColorList_loon()  # default loon palette
l_setColorList(l_colRemoveAlpha(rainbow(5)))  # any set of colours without alpha

To change the palette

• execute one of these functions
• if the inspector is open (not recommended), then select the inspector and close it to force a refresh its colour palette.

l_setColorList_loon() returns the inspector to the default loon colour palette.

accessing the plot

The plot itself is a (tcl) data structure and can be assigned to an R variable.

Either at construction (the preferable way):

# loon graphics (note that the result is assigned to p)
p <- l_plot(x = quakes$long, y = quakes$lat, 
            xlabel = "longitude", ylabel = "latitude",
            title = "Tonga trench earthquakes")

or, if (as we did) you forgot to do that or reassigned something else to that variable (more likely), by accessing the plot from its tcl string:

# accessing the plot from its string representation
p <- l_create_handle(".l0.plot")

Note that this string ".l0.plot" appears on the right side of the window title of an l_plot.

Do this and assign your plot to p. You might have to look at the window to get the string if you have created more than one l_plot().

printing the plot

Were we to now print p

p 

[1] ".l0.plot"
attr(,"class")
[1] "l_plot" "loon" 

we get the string name of the plot, but with a class attribute. This is the value of p and provides us access to a very rich data structure, but one whose values reside in tcl.

For example, we can access this value and, like plots from lattice or ggplot2 produce a grid data structure that can then be displayed as a grid graphic in R.

The simplest way to achieve this is to use the generic R function plot()

plot(p)

Note that this will always reflect the current state of the loon plot. In this way, you can get snapshots of the loon plot after any interaction.

Exercise: Try this. Change p by direct manipulation in loon and repeat.

This allows the visual representation of a loon plot to be displayed at any time.

getting plot contents

There are names associated with a loon plot indicating its contents or states:

head(names(p))

[1] "glyph" "itemLabel" "showItemLabels" "linkingGroup" "linkingKey" "zoomX"

These features of p can be directly accessed the [] function.

For example,

p["showGuides"]

[1] TRUE

A more meaningful example might be identify groups with the colours of the points.
This could be done simply as follows

unique_cols <- unique(p['color'])
group1 <- p['color'] == unique_cols[1]
# Number in group 1 (e.g. as you might have -- here from saved colours)
sum(group1)

[1] 425

# Data on first few quakes in group 1.
head(quakes[group1,])

      lat   long depth mag stations
1  -20.42 181.62   562 4.8       41
2  -20.62 181.03   650 4.2       15
4  -17.97 181.66   626 4.1       19
5  -20.42 181.96   649 4.0       11
10 -17.47 179.59   622 4.3       19
11 -21.44 180.69   583 4.4       13

Alternatively, a function could be written to save the groups as a list of logical vectors (one for each group):

getGroups <- function(loonplot){
    # make sure it is an "l_plot"
    if (!is(loonplot, "l_plot")) stop("loonplot must be an l_plot")
    
    # use color to identify groups
    unique_cols <- unique(loonplot['color'])
    lapply(unique_cols,
           FUN = function(col){loonplot['color'] == col}
           )
}

myGroups <- getGroups(p)    # returns groups identified by unique colour in p
nGroups <- length(myGroups) # number of groups
group1 <- myGroups[[1]]     # each group is an element of the list myGroups

Similarly, other point symbol characteristics (e.g. size, glyph) might be used to define groups. This is particularly handy when these are first determined interactively.

setting plot contents

In addition to accessing the state values of a loon plot using [], its values may also be set using []. For example, try these:

p["showGuides"] <- TRUE

p["size"] <- sample(1:30, size = length(p["x"]), replace = TRUE)

for (i in 1:length(myGroups))  {
    p["selected"] <- myGroups[[i]]
    Sys.sleep(1)
}
p["selected"] <- FALSE

# something a little more involved for up to 6 groups
myCols <- c("firebrick", "steelblue", "purple", 
            "orange", "grey10", "grey80")
for (i in 1:length(myGroups)) {
    p["color"][myGroups[[i]]] <- myCols[i]
}

# something crazy
for (j in 1:10) {
  p["xTemp"] <- p["x"] + runif(length(p["x"]), min = -0.5, max = 0.5)
  Sys.sleep(0.1)
}

# putting locations and size back
p["xTemp"] <- p["x"]
p["size"] <- 4

Example: adding maps

For example, we can add a map to the current loon plot p.

First get the relevant map:

library(maps)
NZFijiMap <- map("world2", regions = c("New Zealand", "Fiji"), plot = FALSE)

It is added as a “layer” to the loon plot

l_layer(p, NZFijiMap, 
        label = "New Zealand and Fiji",
        color = "forestgreen",
        index = "end")
loon layer "New Zealand and Fiji" of type polygons of plot .l0.plot
[1] "layer0"

Try this out.

As can be seen in the world view, much of the map is outside of the plot display (shown as the black-bordered rectangle in the world view).

Now play with the scaling buttons in the loon inspector.

Effect of scaling choices

Adding the map allows us to see the effect of all three plot scaling choices available in the inspector.

The effect is best seen on the world view:

Note that the plot in the world view matches that of the actual plot. Also the aspect ratio changes with the scaling.

the “Layers” tab

The map is added as a layer, which can be seen by selecting the “Layers” tab in the inspector:

This allows a lot of interaction with the layers - for example changing their display order, making some of them invisible, grouping layers, and so on.

Other vignettes

Other vignettes have been developed to cover a range of topics. These currently include:

• An example exploratory analysis: “Visible minorities in Canadian cities” available via

  vignette("minorities", package = "loon")

  This illustrates a variety of loon features including map layers, parallel coordinate axes, and radial axes glyphs in the context of an exploratory analysis of the distribution of visible minorities in 33 Canadian cities.

• New interactive graphics in loon in a teaching example: “Smoothers and Bone Mineral Density” available via

  vignette("teaching-example-smoothing", package = "loon")

  This illustrates how to create new interactions in loon through more general bindings. This is done by developing a teaching tool for explaining how smoothing works. The example data set is bone from the loon.data package.

• Some suggestions on integrating loon plots into an Rmarkdown document: “Loon in RMarkdown” via

  vignette("Rmarkdown", package = "loon")

loon demos

There are numerous demos in loon whose source code might be examined for more examples of using loon. To see them all

demo(package = "loon")  # list all demos

The range is quite broad. For example,

### teaching demos
demo("l_regression", package = "loon")  # lots using the Old Faithful geyser
demo("l_regression_influential", package = "loon")  # change points change the regression fit

### gapminder
demo("l_us_and_them", package = "loon")  #  basic demo
demo("l_us_and_them_slider", package = "loon")  # year selected on a slider
demo("l_us_and_them_choropleth", package = "loon")  # world map and linked with a scatterplot

### the spatial package sp
demo("l_polygons_sp", package = "loon")  # layer polygons with class sp 

### layering and custom layouts
demo("l_layers")  # demonstrate layer types   
demo("l_layout")  # custom layout widgets      
demo("l_widgets")  # inspector and plot in one window

### novel brushing and linking
demo("l_knn")  # brushing by k nearest points in some subspace
demo("l_us_and_them_choropleth")  # many to one linking 

### high dimensional data and dimensionality reduction
demo("l_ng_images_frey_LLE")  # navigation graphs, image data, LLE
demo("l_ng_dimred")  # comparing dimension reduction methods

The code that appears in the console after the demo can be examined and reused to create your own loony loon code.

zenplots

This package is available on CRAN.

install.packages("zenplots")
vignette("intro", package = "zenplots")

Provides a means for compact layout of alternating 1d (e.g. histograms) and 2d (e.g. scatterplots) plots. In will layout base graphics plots, grid graphics plots, and loon plots.

It is similar to pairs plots but allows many many more plots to be displayed. From the introductory vignette:

“A zenplot can show the same information as a pairs plot but with two important display differences.

First, the matrix organization of the pairs layout is replaced by the “zig-zag” layout of zenplot. Second, the number of plots produced is about half that of a pairs plot allowing each plot in a zenplot to be given more visual space.”

Specialty loon extensions - loon.<specialty> packages

There are/will be other packages available which build on loon. Those developed by us are typically prefixed with loon (e.g. loon.ggplot)

These will typically be on CRAN when ready, but development versions are accessible from our great northern diver github repositories. For example,

• https://github.com/great-northern-diver

There you can see related packages as well as what is coming up.