treemap (version 2.4-1)

treemap: Create a treemap

Description

A treemap is a space-filling visualization of hierarchical structures. This function offers great flexibility to draw treemaps. Required is a data.frame (dtf) that contains one or more hierarchical index columns given by index, a column that determines the rectangle area sizes (vSize), and optionally a column that determines the rectangle colors (vColor). The way how rectangles are colored is determined by the argument type.

Usage

treemap(dtf, index, vSize, vColor = NULL, stdErr = NULL, type = "index", fun.aggregate = "sum", title = NA, title.legend = NA, algorithm = "pivotSize", sortID = "-size", mirror.x = FALSE, mirror.y = FALSE, palette = NA, palette.HCL.options = NULL, range = NA, mapping = NA, n = 7, fontsize.title = 14, fontsize.labels = 11, fontsize.legend = 12, fontcolor.labels = NULL, fontface.labels = c("bold", rep("plain", length(index) - 1)), fontfamily.title = "sans", fontfamily.labels = "sans", fontfamily.legend = "sans", border.col = "black", border.lwds = c(length(index) + 1, (length(index) - 1):1), lowerbound.cex.labels = 0.4, inflate.labels = FALSE, bg.labels = NULL, force.print.labels = FALSE, overlap.labels = 0.5, align.labels = c("center", "center"), xmod.labels = 0, ymod.labels = 0, eval.labels = FALSE, position.legend = NULL, drop.unused.levels = TRUE, aspRatio = NA, vp = NULL, draw = TRUE, ...)

Arguments

dtf

a data.frame. Required.

index

vector of column names in dtf that specify the aggregation indices. It could contain only one column name, which results in a treemap without hierarchy. If multiple column names are provided, the first name is the highest aggregation level, the second name the second-highest aggregation level, and so on. Required.

vSize

name of the column in dtf that specifies the sizes of the rectangles. Required.

vColor

name of the column that, in combination with type, determines the colors of the rectangles. The variable can be scaled by the addition of "*" or "/". Note: when omitted for "value" treemaps, a contant value of 1 is taken.

stdErr

name of the column that contains standard errors. These are not used for the treemaps, but only aggregated accordingly and returned as item of the output list.

type

type of the treemap, which determines how the rectangles are colored:

"index":

colors are determined by the index variables. Different branches in the hierarchical tree get different colors. For this type, vColor is not needed.

"value":

the numeric vColor-column is directly mapped to a color palette. This palette is diverging, so that values of 0 are assigned to the mid color (white or yellow), and negative and positive values are assigned to color based on two different hues colors (by default reds for negative and greens for positive values). For more freedom, see "manual".

"comp":

colors indicate change of the vSize-column with respect to the numeric vColor-column in percentages. Note: the negative scale may be different from the positive scale in order to compensate for the ratio distribution.

"dens":

colors indicate density. This is analogous to a population density map where vSize-values are area sizes, vColor-values are populations per area, and colors are computed as densities (i.e. population per squared km).

"depth":

each aggregation level (defined by index) has a distinct color. For this type, vColor is not needed.

"categorical":

vColor is a factor column that determines the color.

"color":

vColor is a vector of colors in the hexadecimal (#RRGGBB) format

"manual":

The numeric vColor-column is directly mapped to a color palette. Both palette and range should be provided. The palette is mapped linearly to the range.

fun.aggregate

aggregation function, only used in "value" treemaps. This function determines how values of the lowest aggregation level are aggregated. By default, it takes the sum. Other sensible functions are mean and weighted.mean. In the latter case, the weights are determined by the vSize variable. Other arguments can be passed on. For weighted.mean, it is possible to assign a variable name for its w argument.

title

title of the treemap.

title.legend

title of the legend.

algorithm

name of the used algorithm: "squarified" or "pivotSize". The squarified treemap algorithm (Bruls et al., 2000) produces good aspect ratios, but ignores the sorting order of the rectangles (sortID). The ordered treemap, pivot-by-size, algorithm (Bederson et al., 2002) takes the sorting order (sortID) into account while aspect ratios are still acceptable.

sortID

name of the variable that determines the order in which the rectangles are placed from top left to bottom right. Only applicable when algorithm=="pivotSize". Also the values "size" and "color" can be used, which refer to vSize and vColor respectively. To inverse the sorting order, use "-" in the prefix. By default, large rectangles are placed top left.

mirror.x

logical that determines whether the rectangles are mirrored horizontally

mirror.y

logical that determines whether the rectangles are mirrored vertically

palette

one of the following:

a color palette:

i.e., a vector of hexadecimal colors (#RRGGBB)

a name of a Brewer palette:

See RColorBrewer::display.brewer.all() for the options. The palette can be reversed by prefixing with a "-". For treemap types "value" and "comp", a diverging palette should be chosen (default="RdYlGn"), for type "dens" a sequential (default="OrRd"). The default value for "depth" is "Set2".

"HCL":

Tree Colors are color schemes derived from the Hue-Chroma-Luminance color space model. This is only applicable for qualitative palettes, which are applied to the treemap types "index", "depth", and "categorical". For "index" and "categorical" this is the default value.

palette.HCL.options

list of advanced options to obtain Tree Colors from the HCL space (when palette="HCL"). This list contains:

hue_start:

number between 0 and 360 that determines the starting hue value (default: 30)

hue_end:

number between hue_start and hue_start + 360 that determines the ending hue value (default: 390)

hue_perm:

boolean that determines whether the colors are permuted such that adjacent levels get more distinguishable colors. If FALSE, then the colors are equally distributed from hue_start to hue_end (default: TRUE)

hue_rev:

boolean that determines whether the colors of even-numbered branched are reversed (to increase discrimination among branches)

hue_fraction:

number between 0 and 1 that determines the fraction of the hue circle that is used for recursive color picking: if 1 then the full hue circle is used, which means that the hue of the colors of lower-level nodes are spread maximally. If 0, then the hue of the colors of lower-level nodes are identical of the hue of their parents. (default: .5)

chroma:

chroma value of colors of the first-level nodes, that are determined by the first index variable (default: 60)

luminance:

luminance value of colors of the first-level nodes, i.e. determined by the first index variable (default: 70)

chroma_slope:

slope value for chroma of the non-first-level nodes. The chroma values for the second-level nodes are chroma+chroma_slope, for the third-level nodes chroma+2*chroma_slope, etc. (default: 5)

luminance_slope:

slope value for luminance of the non-first-level nodes (default: -10)

For "depth" and "categorical" types, only the first two items are used. Use treecolors to experiment with these parameters.

range

range of values (so vector of two) that correspond to the color legend. By default, the range of actual values, determined by vColor, is used. Only applicable for numeric types, i.e. "value", "comp", "dens", and "manual". Note that the range doesn't affect the colors in the treemap itself for "value" and "manual" types; this is controlled by mapping.

mapping

vector of three values that specifies the mapping of the actual values, determined by vColor, to palette. The three values are respectively the minimum value, the mid value, and the maximum value. The mid value is particularly useful for diverging color palettes, where it defined the middle, neutral, color which is typically white or yellow. The mapping should cover the range. By default, for "value" treemaps, it is c(-max(abs(values)), 0, max(abs(values))), where values are the actual values defined by vColor. For "manual" treemaps, the default setting is c(min(values), mean(range(values)), max(values)). A vector of two can also be specified. In that case, the mid value will be the average of those. Only applicable for "value" and "manual" type treemaps.

n

preferred number of categories by which numeric variables are discretized.

fontsize.title

font size of the title

fontsize.labels

font size(s) of the data labels, which is either a single number that specifies the font size for all aggregation levels, or a vector that specifies the font size for each aggregation level. Use value 0 to omit the labels for the corresponding aggregation level.

fontsize.legend

font size for the legend

fontcolor.labels

Specifies the label colors. Either a single color value, or a vector of color values one for each aggregation level. By default, white and black colors are used, depending on the background (bg.labels).

fontface.labels

either a single value, or a vector of values one for each aggregation level. Values can be integers If an integer, following the R base graphics standard: 1 = plain, 2 = bold, 3 = italic, 4 = bold italic, or characters: "plain", "bold", "italic", "oblique", and "bold.italic".

fontfamily.title

font family of the title. Standard values are "serif", "sans", "mono", "symbol". Mapping is device dependent.

fontfamily.labels

font family of the labels in each rectangle. Standard values are "serif", "sans", "mono", "symbol". Mapping is device dependent.

fontfamily.legend

font family of the legend. Standard values are "serif", "sans", "mono", "symbol". Mapping is device dependent.

border.col

color of borders drawn around each rectangle. Either one color for all rectangles or a vector of colors, or one for each aggregation level

border.lwds

thicknesses of border lines. Either one number specifies the line thicknesses (widths) for all rectangles or a vector of line thicknesses for each aggregation level.

lowerbound.cex.labels

multiplier between 0 and 1 that sets the lowerbound for the data label font sizes: 0 means draw all data labels, and 1 means only draw data labels if they fit (given fontsize.labels).

inflate.labels

logical that determines whether data labels are inflated inside the rectangles. If TRUE, fontsize.labels does not determine the fontsize anymore, but it still determines the minimum fontsize in combination with lowerbound.cex.labels.

bg.labels

background color of high aggregation labels. Either a color, or a number between 0 and 255 that determines the transparency of the labels. In the latter case, the color itself is determined by the color of the underlying rectangle. For "value" and "categorical" treemaps, the default is (slightly) transparent grey ("#CCCCCCDC"), and for the other types slightly transparent: 220.

force.print.labels

logical that determines whether data labels are being forced to be printed if they don't fit.

overlap.labels

number between 0 and 1 that determines the tolerance of the overlap between labels. 0 means that labels of lower levels are not printed if higher level labels overlap, 1 means that labels are always printed. In-between values, for instance the default value .5, means that lower level labels are printed if other labels do not overlap with more than .5 times their area size.

align.labels

object that specifies the alignment of the labels. Either a character vector of two values specifying the horizontal alignment ("left", "center", or "right") and the vertical alignment ("top", "center", or "bottom"), or a list of sush character vectors, one for each aggregation level.

xmod.labels

the horizontal position modification of the labels in inches. Either a single value, or a vector that specifies the modification for each aggregation level.

ymod.labels

the vertical position modification of the labels in inches. Either a single value, or a vector that specifies the modification for each aggregation level.

eval.labels

should the text labels, i.e. the factor labels of the index variables, be evaluated as expressions? Useful for printing mathematical symbols or equations.

position.legend

position of the legend: "bottom", "right", or "none". For "categorical" and "index" treemaps, "right" is the default value, for "index" treemap, "none", and for the other types, "bottom".

drop.unused.levels

logical that determines whether unused levels (if any) are shown in the legend. Applicable for "categorical" treemap type.

aspRatio

preferred aspect ratio of the main rectangle, defined by width/height. When set to NA, the available window size is used.

vp

viewport to draw in. By default it is not specified, which means that a new plot is created. Useful when drawing small multiples, or when placing a treemap in a custom grid based plot.

draw

logical that determines whether to draw the treemap.

...

arguments to be passed to other functions. Currently, only fun.aggregate takes optional arguments.

Value

A list is silently returned:

tm

a data.frame containing information about the rectangles: indices, sizes, original color values, derived color values, depth level, position (x0, y0, w, h), and color.

type

argument type

vSize

argument vSize

vColor

argument vColor

stdErr

standard errors

algorithm

argument algorithm

vpCoorX

x-coordinates of the treemap within the whole plot

vpCoorY

y-coordinates of the treemap within the whole plot

aspRatio

aspect ratio of the treemap

range

range of the color values scale

References

Bederson, B., Shneiderman, B., Wattenberg, M. (2002) Ordered and Quantum Treemaps: Making Effective Use of 2D Space to Display Hierarchies. ACM Transactions on Graphics, 21(4): 833-854.

Bruls, D.M., C. Huizing, J.J. van Wijk. Squarified Treemaps. In: W. de Leeuw, R. van Liere (eds.), Data Visualization 2000, Proceedings of the joint Eurographics and IEEE TCVG Symposium on Visualization, 2000, Springer, Vienna, p. 33-42.

Examples

#########################################
### quick example with Gross National Income data
#########################################
data(GNI2014)
treemap(GNI2014,
       index=c("continent", "iso3"),
       vSize="population",
       vColor="GNI",
       type="value")

#########################################
### extended examples with fictive business statistics data
#########################################
data(business)

#########################################
### treemap types
#########################################

# index treemap: colors are determined by the index argument
## Not run: 
# # large example which takes some time...
# treemap(business, 
#         index=c("NACE1", "NACE2", "NACE3"), 
#         vSize="turnover", 
#         type="index")
# ## End(Not run)
treemap(business[business$NACE1=="C - Manufacturing",],
        index=c("NACE2", "NACE3"),
        vSize=c("employees"),
        type="index")

# value treemap: colors are derived from a numeric variable given by vColor 
# (when omited, all values are set to 1 as in the following example)
treemap(business,
        index=c("NACE1", "NACE2"),
        vSize="employees",
        title.legend="number of NACE4 categories",
        type="value")

# comparisson treemaps: colors indicate change of vSize with respect to vColor
treemap(business,
        index=c("NACE1", "NACE2"),
        vSize="employees",
        vColor="employees.prev",
        type="comp")

# density treemaps: colors indicate density (like a population density map)
treemap(business,
        index=c("NACE1", "NACE2"),
        vSize="turnover",
        vColor="employees/1000",
        type="dens")

## Not run: 
# # depth treemap: show depth
# treemap(business,
#         index=c("NACE1", "NACE2", "NACE3"), 
#         vSize="turnover",
#         type="depth")
# ## End(Not run)

# categorical treemap: colors are determined by a categorical variable
business <- transform(business, data.available = factor(!is.na(turnover)), x = 1)
treemap(business,
        index=c("NACE1", "NACE2"),
        vSize="x",
        vColor="data.available",
        type="categorical")

## Not run: 
# # color treemap
# business$color <- rainbow(nlevels(business$NACE2))[business$NACE2]
# treemap(business,
#         index=c("NACE1", "NACE2"), 
#         vSize="x",
#         vColor="color",
#         type="color")
# 
# # manual
# business$color <- rainbow(nlevels(business$NACE2))[business$NACE2]
# treemap(business,
#         index=c("NACE1", "NACE2"), 
#         vSize="turnover",
#         vColor="employees",
#         type="manual",
#         palette=terrain.colors(10))
# ## End(Not run)

#########################################
### graphical options: control fontsizes
#########################################

## Not run: 
# # draw labels of first index at fontsize 12 at the center, 
# # and labels of second index at fontsize 8 top left
# treemap(business, 
#         index=c("NACE1", "NACE2"), 
#         vSize="employees", 
#         fontsize.labels=c(12, 8), 
#         align.labels=list(c("center", "center"), c("left", "top")),
#         lowerbound.cex.labels=1)
#     
#     
# # draw all labels at fontsize 12 (only if they fit)
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         fontsize.labels=12,
#         lowerbound.cex.labels=1)
# 
# # draw all labels at fontsize 12, and if they don't fit, reduce to a minimum of .6*12
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         fontsize.labels=12,
#         lowerbound.cex.labels=.6)
# 
# # draw all labels at maximal fontsize
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         lowerbound.cex.labels=0,
#         inflate.labels = TRUE)
# 
# # draw all labels at fixed fontsize, even if they don't fit
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         fontsize.labels=10,
#         lowerbound.cex.labels=1,
#         force.print.labels=TRUE)
# 
# #########################################
# ### graphical options: color palettes
# #########################################
# 
# ## for comp and value typed treemaps all diverging brewer palettes can be chosen
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         vColor="employees.prev",
#         type="comp",
#         palette="RdBu")
# 
# ## draw warm-colored index treemap
# palette.HCL.options <- list(hue_start=270, hue_end=360+150)
# treemap(business, 
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         type="index",
#         palette.HCL.options=palette.HCL.options)
# 
# # terrain colors
# business$employees.growth <- business$employees - business$employees.prev
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         vColor="employees.growth",
#         type="value",
#         palette=terrain.colors(10))
# 
# # Brewer's Red-White-Grey palette reversed with predefined legend range
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         vColor="employees.growth",
#         type="value",
#         palette="-RdGy",
#         range=c(-20000,30000))
# 
# # More control over the color palette can be achieved with mapping
# treemap(business,
#         index=c("NACE1", "NACE2"),
#         vSize="employees",
#         vColor="employees.growth",
#         type="value",
#         palette="RdYlGn",
#         range=c(-20000,30000),           # this is shown in the legend
#         mapping=c(-30000, 10000, 40000)) # Rd is mapped to -30k, Yl to 10k, and Gn to 40k
# ## End(Not run)