corybrunson
commented
4 years ago Hi @b-tierney, thanks for asking—this is an interesting question, and i don't think there's a natural answer to it yet.
First, is this more or less what you're after? You're right that it's not easy to describe. This is just a workaround; ggalluvial isn't contributing anything. And of course if the basic idea is right then the spacing, size, etc. can be improved.
library(tidyverse)
library(ggalluvial)
demo_data=read.csv('~/demo_alluv_data.txt',row.names=1,stringsAsFactors = FALSE)
demo_data=demo_data %>% group_by(ID) %>% arrange(FREQ)
demo_data$LEVEL1=fct_reorder(factor(demo_data$LEVEL1),demo_data$FREQ)
demo_data$LEVEL2=fct_reorder(factor(demo_data$LEVEL2),demo_data$FREQ)
SUM1 <- with(demo_data, sum(log(FREQ)[! is.na(LEVEL1)]))
SUM2 <- with(demo_data, sum(log(FREQ)[! is.na(LEVEL2)]))
sum_data <- data.frame(axis = 1:2, sum = c(SUM1, SUM2))
ggplot(as.data.frame(demo_data),aes(y = log(FREQ), axis1 = LEVEL1,axis2 = LEVEL2)) +
geom_alluvium(aes(fill = factor(ID)), width = 1/12) +
geom_stratum(width = 1/32, fill = "black", color = "grey") +
geom_label(stat = "stratum", infer.label = TRUE) +
geom_text(data = sum_data, vjust = 0,
aes(x = axis, y = sum, label = round(sum, 2),
axis1 = NULL, axis2 = NULL)) +
scale_fill_brewer(type = "qual", palette = "Set1")
Created on 2020-04-17 by the reprex package (v0.3.0)
Second, do you know whether what you want can be done with geom_bar()? I've found a few similar questions and answers on StackOverflow, but they're also workarounds rather than elegant shortcuts:
- https://stackoverflow.com/q/33026512/4556798
- https://stackoverflow.com/a/46377968/4556798
- https://stackoverflow.com/a/49930429/4556798
The reason i ask is that i've tried to keep ggalluvial consistent with ggplot2 behavior, and in particular the stratum stat and geom similar to the behavior of geom_bar(). If there's a feature missing then i'll definitely want to add it.
b-tierney
Description of the issue
I'm trying to use ggalluvial to create plots for some hierarchical data, for example of the following form (example df also attached):
ID,L1,L2,FREQ A,1,1.1,4 A,1,1.2,5 B,1,1.1,2
However, in these data I sometimes have entries at the higher level (L1 in the above data frame) that are NA -- it's an unavoidable feature of the input data. The result looks like this:
ID,L1,L2,FREQ A,1,1.1,4 A,1,1.2,5 A,1,NA,2
In the example below, you'll see what happens when I try to plot this style of DF with ggalluvial -- the NA values are placed on a non-labeled portion of the L2 axis.
In short, is there anyway to have ggalluvial report the sum by group and category over each column, discounting NAs? Forgive me if that's not clear, it's difficult to explain well.
Thank you in advance.
Reproducible example (preferably using
reprex::reprex())Created on 2020-04-17 by the reprex package (v0.3.0) demo_alluv_data.txt