ggplot2 density-plot with discrete data - r

Question

I want to create a density plot with the following data:
interval fr mi ab
0x 9765 3631 12985
1x 2125 2656 601
2x 1299 2493 191
3x 493 2234 78
4x 141 1559 20
5x and more 75 1325 23
On the X-Axis I want to have the Intervals and on the Y-Axis I want to have the density of "fr", "mi" and "ab" in different colors.
My imagination was something like this graph.
My problem is that I don't know how to get the density on the Y-Axis. I tried it with geom_density, but it didn't work. The best result I accomplished was using the following code:
DS29 <-as.data.frame(DS29)
DS29$interval <- factor(DS29$interval, levels = DS29$interval)
DS29 <- melt (DS29,id=c("interval"))
output$DS51<- renderPlot({
plot_tab6 <- ggplot(DS29, aes(x= interval,y = value, fill=variable, group = variable)) +
geom_col()+
geom_line()
return(plot_tab6)
})
This gives me the following plot, which is not the result I want to have. Do you have an idea how I could get to my wanted result? Thank you very much.

Answer 1

Seeing your sample data, I am not sure if you want to use geom_density. If you type ?geom_density, you will see some example codes. If I take one example from the help page, you may see things that you are missing.
ggplot(diamonds, aes(depth, fill = cut, colour = cut)) +
geom_density(alpha = 0.1) +
xlim(55, 70)
For x-axis, depth is a continuous variable, not a categorical variable. Your current data has a categorical variable in x-axis. For geom_density, you are looking for density of something at a value on x-axis. The example code above shows that the density of diamonds classified as "Ideal" has high density around 61.5-62, suggesting that the largest proportion "Ideal" diamonds have depth value around 61.5-62. Indeed, mean value for depth of "Ideal" diamond is 61.71. This means that you need multiple data points to calculate density. Your data has only one data point for each interval for each group (e.g., ab, fr, mi). So, I do not think your data is not ready for calculating density.
If you want to draw a graphic similar to what you suggested in your question using the current data, I think you need to 1) convert interval to a numeric variable, 2) transform the data into long format, and 3) use stat_smooth.
library(tidyverse)
mydf %>%
mutate(interval = as.numeric(sub(x = as.character(interval), pattern = "x", replacement = ""))) %>%
gather(key = group, value = value, - interval) -> temp
ggplot(temp, aes(x = interval, y = value, fill = group)) +
stat_smooth(geom = "area", span = 0.4, method = "loess", alpha = 0.4)