ggplot: how to get a semi-transparent fill in a geom_rect with a color different from other geoms? fill=alpha() does not seem to work

Question

I am trying to visualize surgical procedures before and after Covid-19:

As you can see, I have a geom_rect() colored differently than the geom_point() .

I want the blue color and fill in the geom_rect() to be semi-transparent, something like fill = alpha("#2C77BF", .5)) . However, when using the script below, the alpha -part does not work.

How can I obtain the semi-transparency of geom_rect() for a color different from those specified for geom_point() ?

ggplot(b,
       aes(x = cons_week, y = n, color = corona, fill = corona)) +
  geom_point(size = 5, shape = 21) +
  geom_smooth(se = F, method = lm, color = "black", show.legend = F) +
  geom_smooth(lty = 2, show.legend = F) + 
  geom_segment(aes(x = 167, xend = 167, y = 2.5, yend = 25), 
               color = "red", size = 1) +
  geom_rect(aes(xmin = 1, xmax = 30, 
                ymin = 0, ymax = 5),
            color = "#2C77BF", 
            fill = alpha("#2C77BF", .5)) +
  scale_color_manual(name = "",
                     values = c("#8B3A62", "#6DBCC3"),
                     labels = c("COVID-19", "Normal"),
                     guide = guide_legend(reverse=TRUE)) + 
  scale_fill_manual(name = "",
                    values = alpha(c("#8B3A62", "#6DBCC3"), .25),
                    labels = c("COVID-19", "Normal"),
                    guide = guide_legend(reverse=TRUE)) + 
  scale_x_continuous(name = "",
                     breaks = seq(0, 210, 12)) + 
  scale_y_continuous(name = "",
                     breaks = seq(0, 30, 5), limits = c(0, 30)) + 
  theme(axis.title.y = element_text(color = "grey20", 
                                    size = 17, 
                                    face="bold", 
                                    margin=ggplot2::margin(r=10)),
        axis.line = element_line(colour = "black"),
        axis.text.x = element_text(color = "white", size = 20),
        axis.ticks.x = element_blank(),
        panel.grid.major = element_line(colour = "grey90"),
        panel.grid.minor = element_line(colour = "grey90"),
        panel.border = element_blank(),
        panel.background = element_blank(),
        legend.position = "top",
        legend.key = element_rect(fill = "white"),
        legend.text=element_text(size=15))

Data

b <- structure(list(corona = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L), .Label = c("C19", "Normal"
), class = "factor"), cons_week = c(185, 176, 190, 201, 184, 
170, 202, 179, 203, 178, 206, 208, 209, 193, 181, 168, 191, 171, 
192, 195, 186, 175, 187, 174, 207, 169, 205, 197, 200, 173, 204, 
199, 189, 180, 194, 188, 182, 177, 196, 183, 124, 111, 75, 148, 
27, 158, 1, 62, 11, 56, 57, 154, 141, 51, 112, 159, 116, 8, 126, 
121, 38, 9, 78, 122, 32, 63, 94, 129, 76, 43, 44, 103, 84, 89, 
92, 37, 67, 19, 73, 142), n = c(17L, 9L, 16L, 15L, 12L, 9L, 15L, 
17L, 11L, 12L, 15L, 14L, 12L, 15L, 13L, 11L, 17L, 14L, 15L, 11L, 
19L, 14L, 16L, 15L, 14L, 13L, 20L, 18L, 9L, 20L, 20L, 18L, 15L, 
10L, 13L, 14L, 21L, 13L, 23L, 18L, 14L, 7L, 14L, 13L, 12L, 14L, 
14L, 16L, 19L, 10L, 14L, 11L, 18L, 12L, 18L, 8L, 10L, 15L, 19L, 
21L, 17L, 11L, 10L, 11L, 14L, 18L, 15L, 13L, 17L, 18L, 15L, 14L, 
17L, 9L, 16L, 15L, 17L, 17L, 13L, 12L)), row.names = c(NA, -80L
), groups = structure(list(corona = structure(1:2, .Label = c("C19", 
"Normal"), class = "factor"), .rows = structure(list(1:40, 41:80), ptype = integer(0), class = c("vctrs_list_of", 
"vctrs_vctr", "list"))), row.names = c(NA, -2L), class = c("tbl_df", 
"tbl", "data.frame"), .drop = TRUE), class = c("grouped_df", 
"tbl_df", "tbl", "data.frame"))

Answer 1

You are using the wrong function. Instead of geom_rect(), use annotate(), and put the alpha parameter inside. You also don't need aes().

(...) +
annotate(geom = "rect", xmin = 1, xmax = 30, 
                ymin = 0, ymax = 5,
            color = "#2C77BF", 
            fill = "#2C77BF", alpha = 0.5) +
(...)

Answer 2

@brunosm's answer is the way to solve this problem. On r2evans 's suggestion, I offer an answer that is more illustrative on the reason why alpha doesn't seem to be working when used with geom_rect() . In truth, you can actually still use geom_rect() , although annotate() is the better option.

The answer lies in the inherent difference between geom_rect() and annotate(geom="rect"... . which is that geoms draw things based on your data= , whereas annotate() does not.

Practically-speaking, it's easiest to see by demonstration. You've already posted that you get a solid box when you use the following code below. Please, note that p in this case is your entire plot code without the geom_rect() part:

p <- # your plot code entirely, but leave out geom_rect()
p + geom_rect(
  aes(xmin = 1, xmax = 30, ymin = 0, ymax = 5),
  color = "#2C77BF", fill = alpha("#2C77BF", .5))

That code gives you a solid blue rectangle and it looks like alpha did not work. In reality, this is not quite the case, and the setting for alpha worked just fine. Let's look at the following counterexample. I'll do the same plot, but this time pass only the first observation in the dataset b to geom_rect() . Note the different result:

p + geom_rect(data=b[1,],
  aes(xmin = 1, xmax = 30, ymin = 0, ymax = 5),
  color = "#2C77BF", fill = alpha("#2C77BF", .5))

It's working as expected! This is because since geom_rect() is a geom , it is basing its draw method on the source data, which is in this case b in its entirety. That means that for every observation in b , geom_rect() is drawing a rectangle with alpha 0.5 . In the end, the dark rectangle you are getting is a result of overplotting - in the case of your dataset, it is the result of drawing 80 rectangles with alpha = 0.5 on top of one another.

For even more illustration, note what happens when I pass two of the observations in b to geom_rect() :

p + geom_rect(data=b[1:2,],
  aes(xmin = 1, xmax = 30, ymin = 0, ymax = 5),
  color = "#2C77BF", fill = alpha("#2C77BF", .5))

Our rectangle is now a bit darker, since ggplot is drawing two rectangles in that same location.

Bottom line here is that annotate() takes the data= source out of the picture. It means you can't annotate something that changes with your data, but if you had that situation, you would use geom_rect() .

I guess if you really want to do it this way with geom_rect() , there's no real problem to do it using geom_rect(data=b[1,],... ... it's just probably not the right way to do it. :)