Marginal densities (or bar plots) on facets in ggplot2 - r

Question

my problem is the following: I have this table below
0 1-5 6-10 11-15 16-20 21-26 27-29
a 0.019 0.300 0.296 0.211 0.117 0.042 0.014
b 0.058 0.448 0.308 0.120 0.042 0.019 0.005
c 0.026 0.277 0.316 0.187 0.105 0.068 0.020
d 0.054 0.297 0.378 0.108 0.108 0.041 0.014
e 0.004 0.252 0.358 0.216 0.102 0.053 0.015
f 0.032 0.097 0.312 0.280 0.161 0.065 0.054
g 0.113 0.500 0.233 0.094 0.043 0.014 0.003
h 0.328 0.460 0.129 0.050 0.020 0.010 0.003
representing some marginal frequencies (by row) for each subgroups of my data (a to h).
My dataset is actually in the long format (very long, counting more than 100 thousand entries), with the first 6 rows as you see below:
RX_SUMM_SURG_PRIM_SITE Nodes.Examined.Class
1 Wedge Resection 1-5
2 Segmental Resection 1-5
3 Lobectomy w/mediastinal LNdissection 6-10
4 Lobectomy w/mediastinal LNdissection 6-10
5 Lobectomy w/mediastinal LNdissection 1-5
6 Lobectomy w/mediastinal LNdissection 11-15
When I plot a barplot by group (the table above is simply the cross tabulation of of these two covariates with the row marginal probabilities taken) here's what happens:
The code I have for this is
ggplot(data.ln.red, aes(x=Nodes.Examined.Class))+geom_bar(aes(x=Nodes.Examined.Class, group=RX_SUMM_SURG_PRIM_SITE))+
facet_grid(RX_SUMM_SURG_PRIM_SITE~.)
Actually I would be very happy only with the marginal frequencies (i.e. the ones in the table) on each y-axis of the facets of the plot (instead of the counts).
Anybody can help me with this?
Thanks for all your help!
EM

Answer 1

geom_bar calculates both counts and proportions of observations. You can access these calculated proportions with either ..prop.. (the old way) or calc(prop) (introduced in newer versions of ggplot2). Use this as your y aesthetic.
You can also get rid of the aes you have in geom_bar, as this is just a repeat of what you've already covered by ggplot and facet_grid.
It looks like your counts/proportions are going to vary widely between groups, so I'm adding free y-scaling to the faceting.
Here's an example of a similar plot with the iris data, which you can model your code off of:
library(tidyverse)
ggplot(iris, aes(x = Sepal.Length, y = calc(prop))) +
geom_bar() +
facet_grid(Species ~ ., scales = "free_y")
Created on 2018-04-06 by the reprex package (v0.2.0).
Edit: the calculated prop variable is proportions within each group, not proportions across all groups, so it works differently when x is a factor. For categorical x, prop treats x as the group; to override this, include group = 0 or some other dummy value in your aes. Sorry I missed that the first time!