r - 在ggplot中保留不同数量的面板的面板大小

Question

我想从不同的数据框中绘制两个图形，然后将它们组合成一个图形(最终图形更复杂)。每个图都显示了两个分类变量(比如“Sex”和“Round”)的数据子集。两个图中绘制的数据类型相同。每个数据集在这些分类变量的级别数上有所不同。

例如，这是绘制在 3 x 2 网格中的模拟数据:

require(ggplot2)
set.seed(10)

# Mock data
N <- 10
rounds <- c("A", "B", "C")
NROUNDS <- length(rounds)

df1 <- data.frame(
    Age = 1:N, 
    Response = rnorm(N*2*NROUNDS), 
    Sex = rep(c("M", "F"), each = N), 
    Round = rep(rounds, each = N*2)
)

# Dimension parameters
panel_width <- 2.5
panel_height <- 1.5
ylims <- c(-4, 4)
units <- "in"
panel_spacing <- 0.1
plot_mar <- 0.25

total_x_margin <- panel_spacing + plot_mar*2
total_y_margin <- panel_spacing*(NROUNDS-1) + plot_mar*2

# Plot the figure
six_panel_plot <- ggplot(df1, aes(x = Age, y = Response)) + 
    geom_line(lwd = 2, color = "#CC79A7") + 
    facet_grid(rows = vars(Round), cols = vars(Sex)) + 
    ylim(ylims) + 
    theme(
        panel.spacing.x = unit(panel_spacing, units),
        panel.spacing.y = unit(panel_spacing, units),
        plot.margin = margin(plot_mar, plot_mar, plot_mar, plot_mar, units)
    ) + theme_bw()

# Save the figure
ggsave("six_panel_plot.png", six_panel_plot, 
    width = total_x_margin + panel_width*2, 
    height = total_y_margin + panel_height*NROUNDS)

我试图根据每个维度的面板数量和边距大小调整图形大小。但是，如果在 4 x 2 网格上创建一个类似的图形，则各个面板的尺寸与之前的图形并不完全相同。

# Mock data
N <- 10
rounds <- c("A", "B", "C", "D")
NROUNDS <- length(rounds)

df2 <- data.frame(
    Age = 1:N, 
    Response = rnorm(N*2*NROUNDS), 
    Sex = rep(c("M", "F"), each = N), 
    Round = rep(rounds, each = N*2)
)


# Dimension parameters
panel_width <- 2.5
panel_height <- 1.5
ylims <- c(-4, 4)
units <- "in"
panel_spacing <- 0.1
plot_mar <- 0.25

total_x_margin <- panel_spacing + plot_mar*2
total_y_margin <- panel_spacing*(NROUNDS-1) + plot_mar*2


eight_panel_plot <- ggplot(df2, aes(x = Age, y = Response)) + 
    geom_line(lwd = 2, color = "#CC79A7") + 
    facet_grid(rows = vars(Round), cols = vars(Sex)) + 
    ylim(ylims) + 
    theme(
        panel.spacing.x = unit(panel_spacing, units),
        panel.spacing.y = unit(panel_spacing, units),
        plot.margin = margin(plot_mar, plot_mar, plot_mar, plot_mar, units)
    ) + theme_bw()

ggsave("eight_panel_plot.png", eight_panel_plot, 
    width = total_x_margin + panel_width*2, 
    height = total_y_margin + panel_height*NROUNDS)

如果我在其他软件(Inkscape、Illustrator 等)中对齐这些图形，则面板的尺寸不同。

如何保持个体大小 面板 跨数字(而不是简单地强制第一个数字有四行)？我想避免使用额外的包，这可能只用 ggplot2 吗？

Answer 1

我没有仅 ggplot 的解决方案，但我将使用 ggplot 本身依赖的一些包(grid & gtable)，因此不需要额外的下载。我假设所需的单位以英寸为单位定义，但很容易更改。

library(grid)

# As per your post
panel_width <- 2.5
panel_height <- 1.5

# Convert plot to gtables
plots <- list(p8 = ggplotGrob(eight_panel_plot), 
              p6 = ggplotGrob(six_panel_plot))

plots <- lapply(plots, function(gt) {
  # Find the positions of panels
  panel_x <- unique(panel_cols(gt)$l)
  panel_y <- unique(panel_rows(gt)$t)
  # Change the sizes of these positions
  gt$widths[panel_x] <- unit(panel_width, "inch")
  gt$heights[panel_y] <- unit(panel_height, "inch")
  gt
})

ggsave("test1.png", plot = plots[[1]])
ggsave("test2.png", plot = plots[[2]])

测试1.png:



测试2.png: