R: how to flip stacked side-by-side barplots in ggplot

Question

> dput(gene_cancer2_f)
structure(list(Gender = structure(c(1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L), .Label = c("Female", "Male"
), class = "factor"), Gene = c("ATM", "ATM", "ATM", "ATM", "Population", 
"Population", "Population", "Population", "ATM", "ATM", "ATM", 
"ATM", "Population", "Population", "Population", "Population"
), Cancer = structure(c(2L, 3L, 5L, 12L, 2L, 3L, 5L, 12L, 2L, 
3L, 5L, 12L, 2L, 3L, 5L, 12L), .Label = c("Brain", "Breast", 
"Colorectal", "Endometrial", "Gastric", "Hepatobiliary", "Kidney", 
"Leukemia", "Melanoma", "Osteosarcoma", "Ovarian", "Pancreatic", 
"Prostate", "Soft Tissue Sarcoma", "Thyroid", "Urinary Bladder"
), class = "factor"), Type = c("RiskToAge70", "RiskToAge70", 
"RiskToAge70", "RiskToAge70", "RiskToAge70", "RiskToAge70", "RiskToAge70", 
"RiskToAge70", "RiskToAge85", "RiskToAge85", "RiskToAge85", "RiskToAge85", 
"RiskToAge85", "RiskToAge85", "RiskToAge85", "RiskToAge85"), 
    Risk = c(21.59862, 3.27479, 1.10073, 1.70754, 8.85253, 1.66318, 
    0.23228, 0.44844, 39.61044, 7.07614, 2.50735, 4.46698, 13.66465, 
    3.59502, 0.52923, 1.17365)), row.names = c(NA, -16L), .Names = c("Gender", 
"Gene", "Cancer", "Type", "Risk"), class = "data.frame")


ggplot(gene_cancer2_f, aes(x = Gene, y = Risk, fill = Type)) +
            geom_bar(stat = 'identity', position = 'stack') + 
            facet_grid(~ Cancer) +
            scale_y_continuous(limits = c(0, 100)) +
            labs(x = "Cancer", y = "Risk %") + 
            guides(fill = guide_legend(title = NULL)) +
            theme_minimal() + 
            theme(plot.title = element_text(size = 12, hjust = 0.5),
                  legend.position = "bottom") +
            scale_fill_manual(values = c("dodgerblue4", "sandybrown"),
                              breaks=c("RiskToAge70", "RiskToAge85"),
                       labels=c(paste("Risk to Age 70", " "), "Risk to Age 85"))

This is the graph I have right now:

But this is what I want the graphs to look like:

I'd like to

1) flip my graph so that the bars are horizontal (I've tried coord_flip, but that didn't work)

2) color-code the side-by-side bars differently (e.g. a different color for "Population,", and of course, the corresponding set of legends (so 4 legends total)

3) get rid of the "ATM" and "Population" labels

sorry didn't read all the way and see your comment so I deleted, working on another angle now — Nate, Apr 04 '18 at 03:25
Try `+ facet_wrap(~ Cancer, ncol = 1) + coord_flip()` instead of `facet_grid(~ Cancer)`. — Maurits Evers, Apr 04 '18 at 03:56

score 3 · Accepted Answer · answered Apr 04 '18 at 04:50

3

We can also use geom_barh from ggstance package

library(ggstance)

ggplot(gene_cancer2_f, aes(x = Risk, y = forcats::fct_reorder(Gene, Risk),
                           group = Gene,
                           fill = Type)) +
  facet_grid(Cancer ~ ., switch = 'y') +
  geom_barh(aes(fill = interaction(Gene, Type)), 
            stat = 'identity', position = 'stackv') +
  scale_x_continuous(limits = c(0, 100)) +
  labs(y = "Cancer", x = "Risk %") + 
  theme_minimal() + 
  theme(
    strip.text.y = element_text(angle = 180),
    axis.text.y=element_blank(),
    axis.ticks.y=element_blank(),
    plot.title = element_text(size = 12, hjust = 0.5),
    legend.position = "bottom") +
  scale_fill_manual("", 
                    values = c("dodgerblue4", "darkgreen",
                               "cornflowerblue", "darkseagreen"),
                    labels = c("ATM Risk to age 70", "ATM Risk to age 85", 
                               "Population Risk to age 70", "Population Risk to age 85"))

answered Apr 04 '18 at 04:50

Tung

26,371
7
91
115

I think there is an option `position_dodgev(height = ...)` in the package but never tried it myself – Tung Apr 04 '18 at 05:07
Thank you. One more question...I've updated my original post with another data.frame (after ###EDIT). I tried running the code in your answer, but somehow the greens and the blues are swapped? Do you happen to know why? – Adrian Apr 04 '18 at 05:33
Because now `ggplot` will plot `Population` before `PTEN` in `GENE`. In your previous data, `ATM` was plotted first. You can control the order by setting `factor/level`. Learn more here http://r4ds.had.co.nz/factors.html – Tung Apr 04 '18 at 05:39
1

Got it. Thank you so much! – Adrian Apr 04 '18 at 05:53
Not a problem. Cheers! – Tung Apr 04 '18 at 05:54

score 1 · Answer 2 · answered Apr 04 '18 at 03:50

Doing a stack and dodge bar chart is outside of the capabilities for ggplot(), see here and typically using facets is the best alternative.

This is a hack to get close-ish.

ggplot(gene_cancer2_f, aes(x = paste(Cancer, Gene), y = Risk, group = Gene, fill = Gene)) +
    geom_col(position = "stack", aes(alpha = Type)) + 
    scale_y_continuous(limits = c(0, 100)) +
    scale_x_discrete(labels = c("Breast", "", "Colo","", "Gas","", "Pan", "")) +
    scale_alpha_manual(values = c(1, .7)) +
    labs(x = "Cancer", y = "Risk %") + 
    coord_flip() +
    theme(axis.ticks.y = element_blank(),
          axis.text.y = element_text(vjust = -2),
          legend.position = "bottom",
          legend.box = "horizontal")

Given that the desire express was to have a cross btwn age and type, I'd suggest you consider using `interaction()` on the factors. — IRTFM, Apr 04 '18 at 03:59

score 1 · Answer 3 · answered Apr 04 '18 at 04:01

You can colour by both Gene and Type using interaction(). This code gets close to the output in your second example, but employs a lot of manipulations along the way, so you should double-check that the visualisation is what you intended.

ggplot(gene_cancer2_f, aes(Gene, Risk)) + 
  geom_col(aes(fill = interaction(Gene, Type, lex.order = TRUE)), 
           position = position_stack(reverse = TRUE)) + 
  facet_grid(Cancer~., switch = "y") + 
  coord_flip() + 
  theme(axis.text.y = element_blank(), 
        axis.ticks.y = element_blank()) + 
  labs(x = "Cancer", 
       y = "Risk of Cancer %") + 
  scale_fill_manual(name = "Risk", 
                    values = c("dodgerblue4", "dodgerblue", 
                               "darkseagreen4", "darkseagreen1"), 
                    labels = c("To age 70 carrier", "To age 85 carrier", 
                               "To age 70 non-carrier", "To age 85 non-carrier"))

R: how to flip stacked side-by-side barplots in ggplot

3 Answers3

Linked