Plot Styling and Aesthetics

This vignette demonstrates how to control the visual aesthetics of the pmxhelpr plotting functions. pmxhelpr styles its plots with ggstylekit, which provides a uniform approach to styling plots.

options(scipen = 999, rmarkdown.html_vignette.check_title = FALSE)
library(pmxhelpr)
library(dplyr, warn.conflicts =  FALSE)
library(ggplot2, warn.conflicts =  FALSE)
library(Hmisc, warn.conflicts = FALSE)
library(patchwork, warn.conflicts = FALSE)

Data and Model Objects

The example datasets used in this vignette are based around the data_sad dataset internal to pmxhelpr. This dataset is based on a single ascending dose (SAD) study of an orally administered drug product with a parallel group food effect (FE) cohort and is formatted in a analysis-ready format for non-linear mixed effects (NLME) modeling. data_sad_pkfit is a modified version of data_sad, with PK model predictions appended.

This vignette will assume familiarity with the structure of these datasets from other vignettes.

data_sad

Dataset definitions can be viewed by calling ?data_sad.

Let’s define some variables that may be useful in plotting and filter down to PK or PD relevant records.

data <- data_sad %>%
  mutate(`Food Status` = ifelse(FOOD == 0, "Fasted", "Fed"),
         DoseFood = paste(DOSE, "mg", `Food Status`)) %>%
  mutate(`Dose and Food` = var_addn(DoseFood, ID))

gof_data <- data_sad_pkfit %>%
  mutate(`Food Status` = ifelse(FOOD == 0, "Fasted", "Fed"),
         DoseFood = paste(DOSE, "mg", `Food Status`)) %>%
  mutate(`Dose and Food` = var_addn(DoseFood, ID))

data_pk <- filter(data, CMT %in% c(1,2))
data_pd <- filter(data, CMT %in% c(1,3))
data_pkfit <- filter(gof_data, CMT %in% c(1,2))
data_nca <- filter(data_sad_nca, PART == "Part 1-SAD")
doseprop_data_nca <- df_doseprop(data_nca, metrics = c("aucinf.obs", "cmax"), metric_name_var = "PPTESTCD")

An example PK model (pkmodel) in mrgmod format is provided in the internal package library. This is loaded using the helper function model_mread_load(), which wraps mrgsolve::mread().

model <- model_mread_load("pkmodel")
#> Building pkmodel_cpp ... done.

Plot Styling with `ggstylekit` in `pmxhelpr`

pmxhelpr re-exports four functions from ggstylekit:

reveal(): map a new data variable to an aesthetic or facet, then re-style
restyle_plot(): adjust a plot style after it has been styled.
combine_styled_plots(): combine styled plots and collect legends
legend_spec(): describe one legend, for the legends style field

Additionally, pmxhelpr includes style preset functions which return a ggstylekit::style_spec() object for each plot_*() function containing the defaults for the plotting family. These functions

style_dvtime()/plot_dvtime(),
style_gof()/plot_gof() ,
style_dvconc()/plot_dvconc(),
style_doseprop()/plot_doseprop(),
style_vpc() / plot_vpc_cont() / plot_vpc_cens()

These style presets are passed to the style argument of plotting functions and use the following syntax:style_<fn>(<map> = c(<role> = <value>)).

Unique roles in pmxhelpr plot families can be passed to per-series maps in the style presets. These elements are overridden based on the roles specified, with remaining roles inheriting the defaults.

plot_dvtime(data_pk, dv_var = ODV, style = style_dvtime())

Basics of Plot Styling with Style Specs

Defaults

Call a style preset function with no arguments to print its defaults.

style_dvtime()
#> <ggstylekit_style_spec>
#>   colors                 NULL
#>   fill                   NULL
#>   linetypes              solid , solid , dashed, dashed
#>   alphas                 0.5, 0.5, 0.0, 1.0, 1.0, 1.0, 1.0
#>   shapes                  1, 16
#>   sizes                  0.75, 1.25
#>   linewidths             0.50, 0.75, 0.75, 0.50, 0.50
#>   point_color            NULL
#>   point_alpha            NULL
#>   point_size             NULL
#>   point_shape            NULL
#>   line_color             NULL
#>   line_alpha             NULL
#>   line_linetype          NULL
#>   line_linewidth         NULL
#>   line_fill              NULL
#>   errorbar_color         NULL
#>   errorbar_alpha         NULL
#>   errorbar_linetype      NULL
#>   errorbar_linewidth     NULL
#>   errorbar_width         NULL
#>   errorbar_fill          NULL
#>   bar_fill               NULL
#>   bar_color              NULL
#>   bar_alpha              NULL
#>   bar_linewidth          NULL
#>   area_fill              NULL
#>   area_color             NULL
#>   area_alpha             NULL
#>   area_linewidth         NULL
#>   box_fill               NULL
#>   box_color              NULL
#>   box_alpha              NULL
#>   box_linewidth          NULL
#>   title                  NULL
#>   xlabel                 NULL
#>   ylabel                 NULL
#>   xlims                  NULL
#>   ylims                  NULL
#>   logx                   NULL
#>   logy                   NULL
#>   xbreaks                NULL
#>   ybreaks                NULL
#>   xminor_breaks          NULL
#>   yminor_breaks          NULL
#>   xtick_labels           NULL
#>   ytick_labels           NULL
#>   xorder                 NULL
#>   yorder                 NULL
#>   equal_axis             NULL
#>   legends                NULL
#>   legend.position        NULL
#>   legend.title.position  top
#>   legend_nrow            NULL
#>   legend_ncol            NULL
#>   legend.title.hjust     NULL
#>   caption_hjust          NULL
#>   fill_alpha             NULL
#>   facet                  NULL
#>   facet_scales           NULL
#>   facet_nrow             NULL
#>   facet_ncol             NULL
#>   theme                  <ggplot2 theme>

Maps within the `style_spec`

The style_spec() object bundles several types of maps:

Per-series maps: apply to multiple plot layers specific to pmxhelpr plot families: colors, fill, linetypes, alphas, shapes, sizes,linewidths
Across-geom maps: apply to an entire geom family: point_color, point_shape, point_size, point_shape, line_color, line_alpha, line_linetype,line_linewidth, line_fill, bar_fill, bar_color, bar_alpha, bar_linewidth, area_fill, area_color, area_alpha, area_linewidth, box_fill, box_color, box_alpha, box_linewidth, fill_alpha
Labels and axes: apply to labels and axes: title, xlabel/ylabel, xlims/ylims, logx/logy, xbreaks/ybreaks, xminor_breaks/yminor_breaks, xtick_labels/ytick_labels, xorder/yorder, equal_axis.
Legends: apply to legends and can be set by a legend_spec() object: legends,legend.position, legend.title.position, legend_nrow/legend_ncol, legend.title.hjust, caption_hjust
Facets: apply to facets: facet, facet_scales, facet_nrow/facet_ncol
Theme: apply a ggplot2 theme object.

pmxhelpr sets a shared house theme based on ggplot2::theme_bw(), with the all minor gridlines and major vertical gridlines removed.

Per-series map Roles by plot family

Unique roles in each plot family style spec are as follows that can be passed to per-series maps are summarized below.

style_dvtime(): Roles include obs_point, obs_line, cent_point, cent_line, cent_errorbar, ref_line, loq_line
style_gof(): Roles include the elements in style_dvtime, plus the overlay colors keyed by label (OBS, DV, PRED, IPRED)
style_dvconc(): Roles include obs_point, ref_line, loess, linear
style_doseprop(): Roles include obs_point, linear
style_vpc(): Roles include obs_point, obs_median_line, obs_pi_line, sim_pi_line, sim_median_line, loq_line, sim_pi_ci, sim_pi_area, sim_median_ci

Available roles in per-series maps can be viewed from the style present object using the $ operator.

style_dvtime()$alphas
#>     obs_point      obs_line    cent_point     cent_line cent_errorbar 
#>           0.5           0.5           0.0           1.0           1.0 
#>      ref_line      loq_line 
#>           1.0           1.0

Builder arguments versus style fields

A few aesthetic decisions belong to the plot builder rather than to the style, because the builder needs them to decide what to draw, not just how to draw it. Where the two overlap, the builder argument wins:

log_y: plot_dvtime(), plot_dvconc(), and plot_gof() set the style logy from the plotting function log_y argument. This is because log_y argument in the plotting function also controls the central tendency summary measure (e.g., arithmetic versus geometric mean).
plot_doseprop() axes and facets: this plotting family applies its own log-log scale_*_log10() and per-metric facet_wrap(), so the style spec’s logx/logy/xbreaks/facet fields are ignored. The facet layout fields (facet_scales, default "free"; facet_nrow; facet_ncol) are honored.
Visibility: plot_gof_shown() / plot_vpc_shown() control which layers are built and the style spec controls how the built layers look. Setting style_vpc(alphas = c(obs_point = 0)) draws an invisible layer and includes it in the legend whereas plot_vpc_shown(obs_point = FALSE) omits it entirely. shown is the preferred method when a layer is not wanted.

Examples

Inline override

A style present can be passed and modified inline within the plotting function call.

plot_dvtime(
  data = data_pk, dv_var = ODV, col_var = `Dose and Food`,
  cent = "mean_sdl", log_y = TRUE, loq_method = 2,
  style = style_dvtime(alphas = c(obs_point = 0))
) +
  scale_x_continuous(breaks = seq(0, 168, 24))+
  labs(y = "Concentration (ng/mL)", x = "Time (hours)")

Reusable style object

Often, one may be generating a series of plots with common styling. In this case, it is advantageous to build a style object once and reuse it across plots. In addition to elements mapped to ggplot2 aesthetic layers, the style spec object can be used to specify axis elements, labels, legends, and themes.

new_dvtime_style <- style_dvtime(
  alphas     = c(obs_point = 0),
  xbreaks = seq(0, 168, 24),
  ybreaks = c(0.1, 1, 10, 100, 1000),
  ylims   = c(0.1, 2000),
  xlabel = "Time (hours)",
  ylabel = "Concentration (ng/mL)",
  title = "Single Ascending Dose"
)

plot_dvtime(data_pk, dv_var = ODV, col_var = `Dose and Food`,
            cent = "mean_sdl", log_y = TRUE, loq_method = 2,
            style = new_dvtime_style)

Error bar cap width

Error bar cap width is the errorbar_width style field, in x-axis units. The presets leave it unset, and plot_dvtime() / plot_gof() then default it to 2.5% of the maximum nominal time so the caps scale with the time axis. Set it in the style to override:

plot_dvtime(data_pk, dv_var = ODV, col_var = `Dose and Food`,
            cent = "mean_sdl", log_y = TRUE, loq_method = 2,
            style = style_dvtime(alphas = c(obs_point = 0), errorbar_width = 10))

Labeling with symbols using `bquote()`

The label keys title, xlabel, and ylabel accept bquote(), which is the practical way to put symbols and subscripts into a label:

plot_dvconc(
  data = data_pd, dv_var = CFB, idv_var = CONC, ref = 0, loess = TRUE,
  style = style_dvconc(
    xlabel = bquote("Concentration" ~ (ng %.% mL^-1)),
    ylabel = bquote(Delta ~ "Response (%)")
  )
)

Styling legends with `legend_spec()`

Overview

The legends map accepts a legend_spec() object or list of legend spec objects. A legend spec describes one legend, keyed either by the data column displayed (col) or by the plot map channel (channel).

plot_dvtime(
  data_pk, dv_var = ODV, col_var = `Food Status`, 
  dose_var = "DOSE", dosenorm = T,
  cent = "median", log_y = TRUE,
  style = style_dvtime(
    ylabel  = "Dose-normalized Concentration (ng/mL)",
    xlabel  = "Time (hours)",
    xbreaks = seq(0, 168, 24),
    legends = legend_spec(
      col = `Food Status`,
      title  = "Meal Status",
      labels = c(Fasted = "Fasted (overnight)", Fed = "Fed (high-fat)")
    ),
    legend.position = "bottom",
    legend_nrow     = 1
  )
)

Other legend_spec() arguments include:

hide = TRUE suppresses a legend while leaving its mapping in place
order sets which legend is drawn first when a plot has multiple
nrow/ncol override the legend_nrow/legend_ncol for the legend specified
combine_with folds a variable into an existing legend, when each key of that legend identifies exactly one value of the folded column.

A legend spec’s labels also relabel matching facet strips, if the facetting is specified in the style spec, rather than added as a new layer with facet_wrap().

plot_dvtime(
  data_pk, dv_var = ODV, col_var = `Food Status`, 
  dose_var = "DOSE", dosenorm = T,
  cent = "median", log_y = TRUE,
  style = style_dvtime(
    ylabel  = "Dose-normalized Concentration (ng/mL/mg)",
    xlabel  = "Time (hours)",
    xbreaks = seq(0, 168, 24),
    legends = legend_spec(
      col = `Food Status`,
      title  = "Meal Status",
      labels = c(Fasted = "Fasted (overnight)", Fed = "Fed (high-fat)")
    ),
    legend.position = "bottom",
    legend_nrow     = 1,
    facet = "Food Status",
    facet_nrow = 1,
    facet_ncol = 2,
    facet_scales = "fixed"
  )
)

The original labels for the variable are retained in facet strips with facet_wrap().

plot_dvtime(
  data_pk, dv_var = ODV, col_var = `Food Status`, 
  dose_var = "DOSE", dosenorm = T,
  cent = "median", log_y = TRUE,
  style = style_dvtime(
    ylabel  = "Dose-normalized Concentration (ng/mL/mg)",
    xlabel  = "Time (hours)",
    xbreaks = seq(0, 168, 24),
    legends = legend_spec(
      col = `Food Status`,
      title  = "Meal Status",
      labels = c(Fasted = "Fasted (overnight)", Fed = "Fed (high-fat)")
    ),
    legend.position = "bottom",
    legend_nrow     = 1
  )
)  + 
  facet_wrap(~`Food Status`)

Remove a legend

To drop a legend entirely, one can key the legend spec to the channel.

plot_dvtime(
  data_pk, dv_var = ODV, col_var = `Dose and Food`, cent = "none", id_var = ID,
  log_y = TRUE, 
  style = style_dvtime(
    ylabel  = "Concentration (ng/mL)",
    xlabel  = "Time (hours)",
    xbreaks = seq(0, 168, 24),
    legends = legend_spec(channel = "color", hide = TRUE)
  ) 
) + 
  facet_wrap(~`Dose and Food`)

Per-series maps in `pmxhelpr` style presets

Palette functions

A per-series map can be a palette function(n) rather than a named vector. ggstylekit calls the palette function with the unique number of groups in each plot, so one house style colors every plot without naming its groups.

viridis_pal <- function(n) grDevices::hcl.colors(n, "Viridis")

plot_dvtime(
  data_pk, dv_var = ODV, col_var = `Dose and Food`, cent = "mean", log_y = TRUE,
  style = style_dvtime(colors = viridis_pal,
                       ylabel = "Concentration (ng/mL)", xlabel = "Time (hours)",
                       xbreaks = seq(0, 168, 24))
) + 
  facet_wrap(~PART)

Any function(n) returning n values works. A palette has no names to merge entry-wise, so it replaces the preset default map rather than merging onto it.

Trend line customization (`plot_dvconc()`)

Trend line colors are controlled using the per-series map colors with the loess/linear roles. The 95% CI ribbon is controlled by the ggstylekit default line_fill.

plot_dvconc(
  data = data_pd, dv_var = CFB, idv_var = CONC, ref = 0,
  loess = TRUE, se_loess = TRUE,
  style = style_dvconc(colors = c(loess = "darkred"), line_fill = "darkred")
) +
  labs(x = "Concentration (ng/mL)", y = "Response (% CFB)")

Dose-proportionality facet layout (`plot_doseprop()`)

plot_doseprop() facets by exposure metric, so the style’s facet_ncol, facet_nrow, and facet_scales fields control that layout. Here the two metrics are stacked in a single column.

plot_doseprop(dplyr::filter(data_sad_nca, PART == "Part 1-SAD"),
              metrics = c("aucinf.obs", "cmax"),
              style = style_doseprop(colors = c(linear = "navy"),
                                     facet_ncol = 1))

GOF overlay colors (`plot_gof()`)

The DV (OBS), PRED, and PRED central tendency lines are the role keys for the per-series map colors.

plot_gof(
  data = data_pkfit, dv_var = ODV, cent = "mean", log_y = TRUE,
  style = style_gof(colors = c(OBS = "grey60", DV = "black",
                               PRED = "#3388cc", IPRED = "firebrick"), 
                    xlabel = "Time (hours)", ylabel = "Concentration (ng/mL)", 
                    xbreaks = seq(0, 168, 24)))

VPC plot color scheme (`plot_vpc_cont()`)

The per-series maps colors controls colors for points and lines while confidence/prediction interval ribbon fills are controlled by fill.

sim <- df_mrgsim_replicate(data = data_pk, model = model, replicates = 100,
                           dv_var = ODV, carry_out = c("LLOQ", "FOOD"))

pcvpc_style <- style_vpc(
  colors = c(obs_point = "#000000", obs_median_line = "#000000",
             obs_pi_line = "#000000"),
  fill   = c(sim_median_ci = "#3388cc", sim_pi_ci = "#3388cc"),
  ylabel = "PRED-corrected Conc. (ng/mL)",
  xlabel = "Time (hours)",
  xbreaks = seq(0, 168, 24),
  xlims = c(0, 168)
)

plot_vpc_cont(sim, loq = 1, pcvpc = TRUE,
              style = pcvpc_style) + 
  scale_y_log10(guide = "axis_logticks")

Faceted VPC plot with free scales (`plot_vpc_cont()`)

As discussed in Visual Predictive Check Workflow, VPC plots should ONLY be facetted via the strat_var argument and SHOULD NOT be facetted post-hoc with facet_wrap. This is to ensure that the summary statistics from the simulation are also calculated stratified by the faceting variable.

However, because the facet_wrap call is internal to the VPC plotting functions, the user is not able to change the scales argument from the "fixed" preset to "free_y". This can be accomplished in the style.

sim <- df_mrgsim_replicate(data = data_pk, model = model, replicates = 100,
                           dv_var = ODV, carry_out = c("LLOQ", "FOOD", "DOSE"))

vpc_style_facet <- style_vpc(
  ylabel = "Concentration (ng/mL)",
  xlabel = "Time (hours)",
  xbreaks = seq(0, 168, 24),
  xlims = c(0, 168), 
  facet_scales = "free_y"
)

plot_vpc_cont(sim, loq = 1, strat_var = DOSE,
              style = vpc_style_facet)

VPC legend (`plot_vpc_legend()`)

The VPC plotting functions do not generate a legend. plot_vpc_legend() builds a standalone legend plot intended to be combined with the VPC plot into a single figure.

Pass the same style object along with the same shown, ci, and pi as the plotting function to generate the corresponding legend.

plot_vpc_legend(style = pcvpc_style, lloq = 1)

Updating the plot theme

The theme field takes any ggplot2 theme object and replaces the pmxhelpr house theme wholesale. Build on the house theme by adding to it rather than starting over:

plot_dvtime(
  data_pk, dv_var = ODV, col_var = `Dose and Food`,
            cent = "mean_sdl_upper",
  style = style_dvtime(
    ylabel = "Concentration (ng/mL)", xlabel = "Time (hours)",
    xbreaks = seq(0, 168, 24),
    theme  = theme_minimal(base_size = 8) + 
      theme(panel.grid.minor = element_blank(),
                   panel.grid.major = element_blank(),
                   legend.position  = "bottom")
  )
)

Post-hoc styling: making aesthetic changes with `restyle_plot()`

ggstylekit not only handles plot styling a priori, it also handles plot styling a posteriori after a finished plot object has been generated.

restyle_plot() merges style fields onto a finished plot. Restyling works analogously to the up front styling with style specs.

For example, if one wanted to remove points and the legend from an existing plot, this can be accomplished as follows with restyle_plot.

p <- plot_dvtime(data_pk, dv_var = ODV, cent = "median", log_y = TRUE, 
                 loq_method = 2,
                 style = style_dvtime(
                   xlabel = "Time (hours)", 
                   ylabel = "Dose-normalized Conc. (ng/mL/mg)",
                   xbreaks = seq(0, 168, 24)
                 ))
p


restyle_plot(p, 
             alphas = c(obs_point = 0), 
             legends = legend_spec(channel = "linetype", hide = TRUE))

Post-hoc styling: surfacing new variables with `reveal()`

One of the most powerful features of ggstylekit is the reveal() function.

reveal() maps a variable present in the underlying plot dataset that is not currently mapped to a graphical channel (or a facet).

Mapping to all layers with `as =`

Use as = to inherit that channel’s default values.

reveal(p, `Food Status`, as = "colors")

Alternatively, the values can be supplied manually.

reveal(p, `Food Status`,
       colors = c(Fasted = "#3388cc", Fed = "firebrick"),
       labels = c(Fasted = "Fasted (overnight)", Fed = "Fed (high-fat)"))

Facets are also a channel that can be mapped like any other with reveal.

p2 <- reveal(p, `Food Status`, as = "facet", 
       colors = c(Fasted = "#3388cc", Fed = "firebrick"),
       labels = c(Fasted = "Fasted (overnight)", Fed = "Fed (high-fat)"))

p2

Targeting specific layers with `on`

By default, reveal() maps the variable onto every layer that can display the channel. The on argument narrows this to an entity — "point", "line", "bar", "area", or "box" — or to a named series.

Here the central tendency line and its error bars are left unmapped, reflecting the totality of the data, while the points are mapped to food status with on

p_sdl <- plot_dvtime(data_pk, dv_var = ODV, cent = "mean_sdl", log_y = TRUE,
                     style = style_dvtime(xlabel = "Time (hours)",
                                          ylabel = "Concentration (ng/mL)",
                                          xbreaks = seq(0, 168, 24)))

reveal(p_sdl, `Food Status`, as = "colors", on = "point")

Values are supplied with on exactly as they are without it.

reveal(p_sdl, `Food Status`, on = "point",
       colors = c(Fasted = "#3388cc", Fed = "firebrick"),
       labels = c(Fasted = "Fasted (overnight)", Fed = "Fed (high-fat)"))

on is also how a channel that is already in use is revealed onto. A plot built with col_var already maps color, and revealing a second variable onto color without on returns the plot unchanged, with a message naming the entities and series available to target. Two variables sharing one channel are drawn in a single legend, so revealing onto a free channel — as = "shapes", as = "linetypes", as = "facet" — is usually the clearer choice.

Combining plots with `combine_styled_plots`

combine_styled_plots() arranges several styled plots into one figure with a single collected legend.

Build the panels with a shared style so the collected legend is meaningful.

panel_style_gm <- style_dvtime(
  alphas = c(obs_point = 0.3),
  xlabel = "Time (hours)",
  ylabel = "Concentration (ng/mL)",
  title = "Geometric Mean"
)

panel_style_med <- style_dvtime(
  alphas = c(obs_point = 0.3),
  xlabel = "Time (hours)",
  ylabel = "Concentration (ng/mL)",
  title = "Median"
)

p_mean <- plot_dvtime(data_pk, dv_var = ODV, col_var = `Food Status`,
                      cent = "mean", log_y = TRUE,
                      style = panel_style_gm)

p_median <- plot_dvtime(data_pk, dv_var = ODV, col_var = `Food Status`,
                        cent = "median", log_y = TRUE,
                        style = panel_style_med)

Then combine them. Unnamed arguments are the plots. Named arguments pass through to patchwork::wrap_plots(), so axes = "collect" also de-duplicates the shared axis labels.

p_combo <- combine_styled_plots(p_mean, p_median, ncol = 2, axes = "collect")
p_combo

The combined figure is still a styled object, so restyle_plot() and reveal() continue to work. Here the collected legend is re-titled and moved to the bottom of the figure:

restyle_plot(p_combo, 
             legends = legend_spec(`Food Status`, title = "Meal Status",
                                   labels = c(Fasted = "Fasted (overnight)", 
                                              Fed = "Fed (high-fat)")),
             colors = c(Fasted = "#3388cc", Fed = "firebrick"),
             legend.position = "bottom")

legend.position moves the collected legend on the combined figure, as restyle_plot() reaches the assembled figure and not only the individual panels. Note that labels is not a style field: legend entries are relabeled through the labels argument of legend_spec(), as above.

Panels do not have to come from the same plot family; however, legends need to be identical and drawn from the same layer sets. Here concentration vs time and response vs concentration panels colored by an indicator variable for female sex (SEXF).

The legend is suppressed in one plot in order to have only a single legend collected into the paneled combined plot, as even though the legends are visually identical, they are built from different layers.

p_pk <- plot_dvtime(data_pk, dv_var = ODV, col_var = SEXF,
                    dose_var = DOSE, dosenorm = TRUE,
                    cent = "mean", log_y = TRUE,
                    style = style_dvtime(xlabel = "Time (hours)",
                                         ylabel = "Dose-normalized Conc. (ng/mL/mg)", 
                                         xbreaks = seq(0, 168, 24),
                                         legends = legend_spec(channel = "color", 
                                                               hide = TRUE)))

p_pd <- plot_dvconc(data_pd, dv_var = CFB, idv_var = CONC, col_var = SEXF,
                    col_trend = TRUE, ref = 0, loess = TRUE,
                    style = style_dvconc(xlabel = "Concentration (ng/mL)",
                                         ylabel = "Response (% CFB)"))

combine_styled_plots(p_pk, p_pd, ncol = 2)

Two things to keep in mind when generated combined plots with combine_styled_plots:

combine_styled_plots() flattens the input plot panels and discards any patchwork annotations or layouts. Add titles and tags to the combined plot returned with patchwork::plot_annotation().
guides = "collect" is the default. Pass guides explicitly to override it.

combine_styled_plots(p_pk, p_pd, ncol = 2) +
  plot_annotation(title = "Single ascending dose PK/PD", tag_levels = "A")

Data and Model Objects

data_sad

Plot Styling with ggstylekit in pmxhelpr

Basics of Plot Styling with Style Specs

Defaults

Maps within the style_spec

Per-series map Roles by plot family

Builder arguments versus style fields

Examples

Inline override

Reusable style object

Error bar cap width

Labeling with symbols using bquote()

Styling legends with legend_spec()

Overview

Label overrides in legends and facets

Remove a legend

Per-series maps in pmxhelpr style presets

Palette functions

Trend line customization (plot_dvconc())

Dose-proportionality facet layout (plot_doseprop())

GOF overlay colors (plot_gof())

VPC plot color scheme (plot_vpc_cont())

Faceted VPC plot with free scales (plot_vpc_cont())

VPC legend (plot_vpc_legend())