Customize plots

Benjamin Guiastrennec

29 June, 2021

Plot type

The option type allows to control how the data should be represented.

Scatter plots

In scatter plots type can be any combination of: points 'p', lines 'l', smooth 's', text 't'.

Distribution plots

In distribution plots type can be any combination of: histogram 'h', density 'd', or rug 'r'.

VPCs

In visual predictive checks plots type can be any combination of: areas 'a', lines 'l', points 'p', text 't' or rug 'r'.

Visual guide

The option guide can enable (TRUE) or disable (FALSE) the visual guide on the graphs such as the line of identity on the dv_vs_ipred() type plot or horizontal lines on residual plots (e.g. res_vs_idv()).

Labels

All xpose plots have by default an informative title, subtitle and caption. For example all plots using individual model predictions (IPRED) will display the epsilons’ shrinkage. These titles can easily be edited as templates using @keywords which will be replaced by their actual value stored in the summary level of the xpdb object when rendering the plots. Keywords are defined by a word preceded by a @ e.g. '@ofv'. A list of all available keyword can be accessed via help('template_titles'). The title, subtitle or caption can be disabled by setting them to NULL. Suffix can be automatically added to title, subtitle and caption of all plots. The suffixes can be defined in the xp_theme.

From the plot functions

# Using template titles
dv_vs_ipred(xpdb,
            title    = '@y vs. @x (@run, obj: @ofv)',
            subtitle = 'Based on: @nind subjects and @nobs records',
            caption  = 'Dir: @dir')

# Disabling all titles
dv_vs_ipred(xpdb, title = NULL, subtitle = NULL, caption = NULL)

# Edit title suffix from the xp_theme for a specific plot
dv_vs_ipred(xpdb, title = 'A title', xp_theme = list(title_suffix = ' | a suffix for @run'))

Using the labs function

Labels can also modified later on by using the ggplot2::labs() function in combination with the ggplot2 + operator.

dv_vs_ipred(xpdb) +
  labs(title    = '@descr',
       subtitle = NULL,
       caption  = 'Based on @nobs observations',
       x        = 'Individual model predictions (@x)',
       y        = 'Observations (@y) for @nind subjects')

Modify aesthetics

By default the aesthetics are read from the xp_theme level in the xpdb object but these can be modified in any plot function. xpose makes use of the ggplot2 functions mapping for any layer (e.g. points, lines, etc.) however to direct the mapping to a specific layer, a prefix appealing to the targeted layer should be used. The format is defined as layer_aesthetic = value. Hence to change the color of points in ggplot2 the argument color = 'green' could be used in geom_point(), while in xpose the same could be achieved with point_color = 'green'.

In basic goodness-of-fit plots, the layers have been named as: point_xxx, line_xxx, smooth_xxx, guide_xxx, xscale_xxx, yscale_xxx where xxx can be any option available in the ggplot2 layers: geom_point, geom_line, geom_smooth, geom_abline, scale_x_continuous, etc.

dv_vs_ipred(xpdb, 
            # Change points aesthetics
            point_color = 'blue', point_alpha = 0.5, 
            point_stroke = 0, point_size = 1.5, 
            # Change lines aesthetics 
            line_alpha = 0.5, line_size = 0.5, 
            line_color = 'orange', line_linetype = 'solid', 
            # Change smooth aesthetics
            smooth_method = 'lm')

Aesthetics can also be defined mapped to a variable using the ggplot2 aes() function.

dv_vs_ipred(xpdb, type = 'p', aes(point_color = SEX))

Grouping variable

The argument group defines the grouping variable to be used in the dataset for the lines only, by default group is defined to the column 'ID'. To apply a grouping variable on any other layer, a manual mapping can be made using the argument xxx_group = variable

Panels

Panels (or faceting) can be created by using the facets argument as follows:

# Example with a string
dv_vs_ipred(xpdb, facets = c('SEX', 'MED1'), ncol = 2, nrow = 1, page = 1)

# Example with a formula
dv_vs_ipred(xpdb, facets = SEX~MED1, margins = TRUE)

All xpose plot functions accept arguments for facet_wrap_pagninate and facet_grid_paginate (e.g. ncol = 2, labeller = 'label_both', etc.). With the default xpose theme scales are set to 'free' from one panel to another (scales = 'free'), this behavior can be changed with scales = 'fixed', 'free_y' or 'free_x'.

For more information on the available faceting options read the help of facet_wrap_paginate and facet_grid_paginate from the package ggforce.

Additional layers

xpose offers the opportunity to add any additional layers from ggplot2. Hence, a ggplot2::geom_rug() layer could be added to the dv_vs_ipred() plot along with some annotations (ggplot2::annotate()). Note: the additional layers do not inherit from the xpose aesthetic mapping (i.e. colors or other options need to be defined in each layer as shown below).

Layers can also be used to modify the aesthetics scales for example ggplot2::scale_color_manual(), or remove a legend ggplot2::scale_fill_identity().

dv_vs_ipred(xpdb) +
  geom_rug(alpha = 0.2, color = 'grey50',
           sides = 'lb', size = 0.4) +
  annotate(geom = 'text',
           fontface = 'bold',
           color = 'darkred',
           size  = 3,
           label = 'Toxic concentrations', x = 1.35, y = 1.75) +
  annotate(geom = 'rect',
           alpha = 0.2, fill = 'red',
           xmin = 1, xmax = Inf,
           ymin = 1, ymax = Inf)

Scales options

The argument log allows to log-transform the axes. Accepted values are x, y or xy.

dv_vs_ipred(xpdb, log = 'xy', subtitle = 'Plot on log scale')

Additional arguments can be provided to the scales via the mapping by using the naming convention xscale_xxx or yscale_xxx where xxx is the name of a ggplot2 scale argument such as name, breaks, labels, expand.

dv_vs_ipred(xpdb, 
            xscale_breaks = c(0.3, 0.8, 1.3),
            xscale_labels = c('Low', 'Med', 'High'),
            xscale_expand = c(0.2, 0),
            xscale_name = 'Individual model predictions')

Themes

Theme in xpose are easily customizable. Themes are made up of two parts:

Themes can be attached to an xpdb when importing the data with xpose_data() or using the function update_themes() on an xpdb object.

# While creating the xpdb
xpdb <- xpose_data(runno = '001', 
                   gg_theme = theme_minimal(), 
                   xp_theme = theme_xp_xpose4())

# Update a pre-existing xpdb
xpdb <- update_themes(xpdb     = xpdb,
                      gg_theme = theme_bw2(),
                      xp_theme = list(point_color = 'dodgerblue4',
                                      line_color  = 'dodgerblue4')) 

Examples of gg_theme:

xpose brings two additional themes (theme_readable() and theme_bw2()), to those available in ggplot2.

Examples of xp_theme: