Skip to content

Missing shapes in legend with scale_shape_manual

2 messages · Kevin Zembower, Rui Barradas

Original
Kevin Zembower
# 
Hello,

I'm trying to plot a graph of blood glucose versus date. I also record
conditions, such as missing the previous night's medications, and
missing exercise on the previous day. My data looks like:
# A tibble: 7 ? 5
  Date       Time      bg missed_meds no_exercise
  <date>     <time> <dbl> <lgl>       <lgl>      
1 2023-10-17 08:50    128 TRUE        FALSE      
2 2023-10-16 06:58    144 FALSE       FALSE      
3 2023-10-15 09:17    137 FALSE       TRUE       
4 2023-10-14 09:04    115 FALSE       FALSE      
5 2023-10-13 08:44    136 FALSE       TRUE       
6 2023-10-12 08:55    122 FALSE       TRUE       
7 2023-10-11 07:55    150 TRUE        TRUE
This gets me most of the way to what I want:

ggplot(data = b2, aes(x = Date, y = bg)) +
    geom_line() +
    geom_point(data = filter(b2, missed_meds),
               shape = 20,
               size = 3) +
    geom_point(data = filter(b2, no_exercise),
               shape = 4,
               size = 3) +
    geom_point(aes(x = Date, y = bg, shape = missed_meds),
               alpha = 0) + #Invisible point layer for shape mapping
    scale_y_continuous(name = "Blood glucose (mg/dL)",
                       breaks = seq(100, 230, by = 20)
                       ) +
    geom_hline(yintercept = 130) +
    scale_shape_manual(name = "Conditions",
                       labels = c("Missed meds",
                                  "Missed exercise"),
                       values = c(20, 4),
                       ## size = 3
                       )

However, the legend just prints an empty square in front of the labels.
What I want is a filled circle (shape 20) in front of "Missed meds" and
a filled circle (shape 4) in front of "Missed exercise."

My questions are:
 1. How can I fix my plot to show the shapes in the legend?
 2. Can my overall plotting method be improved? Would you do it this
way?

Thanks so much for your advice and guidance.

-Kevin
Rui Barradas
# 
?s 20:55 de 30/10/2023, Kevin Zembower via R-help escreveu:
Hello,

In ggplot2 graphics when you have more than one call to the same layer 
function, then you can probably simplify the code.

In this case you make several calls to geom_point. This can probably be 
avoided.

Create a new column named Condition.
Assign to it the column names wherever the values of those columns are 
TRUE. The simplest way of doing this is to use colus missed_meds and 
no_exercise as logical index columns, see code below.

Like this the values are mapped to shapes in just one call to geom_point.
That's what function aes() is meant for, to tell what variables define 
what in the plot.



b2$Date <- as.Date(b2$Date)
# this new column will be mapped to the shape aesthetic
b2$Conditions <- NA_character_
b2$Conditions[b2$missed_meds] <- names(b2)[4]
b2$Conditions[b2$no_exercise] <- names(b2)[5]

ggplot(data = b2, aes(x = Date, y = bg)) +
   geom_line() +
   geom_point(aes(shape = Conditions), size = 3) +
   geom_hline(yintercept = 130) +
   scale_y_continuous(
     name = "Blood glucose (mg/dL)",
     breaks = seq(100, 230, by = 20)
   ) +
   scale_shape_manual(
     #name = "Conditions",
     labels = c("Missed meds", "Missed exercise"),
     values = c(20, 4),
     na.translate = FALSE
   )



Hope this helps,

Rui Barradas