如何在ggplot2中的汇总数据上创建小提琴图？

Question

I'm trying to create a violin plot in ggplot2 on aggregated data which consists of pre-calculated counts (the raw data is too big and takes really long to plot). Example dataset below.

data <- data.frame(category = rep(LETTERS[1:3],3),
               value = c(1,1,1,2,2,2,3,3,3),
               count = c(3,2,1,1,2,3,2,1,3))

The solution found here Violin Plot (geom_violin) with aggregated values doesn't seem to work properly, giving inaccurate densities. When I plot using that method and compare against plotting with the entire dataset, the plots looks very different.

Anyone know how to plot on aggregated data?

Answer 1

Any difference you're seeing might be caused simply by the difference in how the density is being calculated. Here is the example you gave:

data <- data.table(category = rep(LETTERS[1:3],3),
                   value = c(1,1,1,2,2,2,3,3,3),
                   count = c(3,2,1,1,2,3,2,1,3))
data[, count2 := count/sum(count), by = category]
ggplot(data, aes(x = category, y = value, weight = count2)) + geom_violin()

Now here is the same data with the values repeated, rather than counts:

val2 <- unlist(sapply(1:length(data$value), 
                      function(x) rep(data$value[x], 
                                      data$count[x])))
cat2 <- unlist(sapply(1:length(data$value), 
                      function(x) rep(data$category[x], 
                                      data$count[x])))

dat2 <- data.table(cat2, val2)
ggplot(dat2, aes(x = cat2, y = val2)) + geom_violin()

This generates the plot:

Not a lot of difference, and likely caused by how the density is being calculated.