dplyr: categorical counts from a single column across multiple variables - r

Question

So I have been trying to do a boxplot of "yes/no" counts for hours now.
My dataset looks like this
> stack
Site Plot Treatment Meters Retrieved
2 Southern 18 Control -5.00 y
3 Southern 18 Control 9.55 y
4 Southern 18 Control 4.70 y
5 Southern 27 Control -5.00 y
6 Southern 27 Control 20.00 n
9 Southern 18 Control -0.10 y
17 Southern 18 Control 20.00 y
23 Southern 31 Control 100.00 y
53 Southern 25 Mu 3.55 n
54 Southern 20 Mu 5.90 y
55 Southern 25 Mu -0.10 y
56 Southern 29 Mu 9.55 y
58 Southern 25 Mu 4.70 y
60 Southern 20 Mu 2.90 y
61 Southern 24 Mu 5.90 n
62 Southern 24 Mu 3.55 y
63 Southern 20 Mu 3.55 y
65 Southern 24 Mu 0.55 y
66 Southern 29 Mu 8.90 y
68 Southern 25 Mu 8.90 y
69 Southern 29 Mu 0.55 y
70 Southern 24 Mu 1.70 y
72 Southern 29 Mu -5.00 y
76 Southern 29 Mu 1.70 y
77 Southern 25 Mu 9.55 y
78 Southern 25 Mu 13.20 y
79 Southern 29 Mu 3.55 y
80 Southern 25 Mu 15.00 y
81 Southern 25 Mu -5.00 n
84 Southern 24 Mu 8.90 y
85 Southern 20 Mu 6.55 y
86 Southern 29 Mu 2.90 y
92 Southern 24 Mu -0.10 y
93 Southern 20 Mu 100.00 y
I want to get counts of both y(yes) and n(no) of the variable "Retrieved" while grouping for "Treatment" and "Meters".
So it should look something like this
Treatment Meters Yes No
Control -5.00 2 0
Control 9.55 1 2
Control 4.70 1 1
Control 20.00 0 2
Mu 3.55 4 0
Mu 5.90 0 1
Mu -0.10 2 2
Mu 9.55 1 0
With this data I want to do a stacked boxplot with x=Meters, y= count and Treatment as grid or something. like this
This is my code but it's not working
plot_data <- stack %>%
count(Retrieved, Treatment, Meters) %>%
group_by(Treatment, Meters) %>%
mutate(count= n)
plot_data
ggplot(plot_data, aes(x = Meters, y = count, fill = Treatment)) +
geom_col(position = "fill") +
geom_label(aes(label = count(count)), position = "fill", color = "white", vjust = 1, show.legend = FALSE) +
scale_y_continuous(labels = count)
Could you please tell me what I'm doing wrong.

Answer 1

geom_bar is for precisely this case, and you won't even need to use group_by or count. (From the docs: "geom_bar makes the height of the bar proportional to the number of cases in each group".)
This should do what you're looking for:
ggplot(stack, aes(x = Meters, fill = Treatment)) +
geom_bar(position = "stack")
However, the bars will be very narrow because "Meters" is continuous and has a large range. You could address this by converting it into a factor. One way to do that would be to do this first:
data <- data %>%
mutate(Meters = as.factor(Meters))
If you want to get the counts in the format that you mentioned (in addition to creating the plot), you could do:
data %>%
count(Treatment, Meters, Retrieved) %>%
spread(Retrieved, n, fill = 0) %>%
rename(Yes = y, No = n)
count does group_by for you, so I didn't need to carry that over from your code. Then, spread creates the separate columns for y and n. Finally, I rename those columns to Yes and No.