使用ggplot2用线图覆盖条形图

Question

My question is similar to those posted here and here .

I am working on creating a graph in ggplot where I have one bar plot and then want to overlay multiple line graphs. For the purposes of this question, I have reproduced my code for two barplots (one that includes all years (2007-2015) and two from specific years (2007 and 2015), but ultimately I will be overlaying data from 10 different years. The data used can be found here .

library(dplyr)
library(tidyr)
library(gridExtra)
library(ggplot2)

overallpierc<-data[(data$item=="piercing"),]

overp<-overallpierc %>%
  group_by(age) %>% 
  count(sex) %>% 
  ungroup %>% 
  mutate(age = factor(age)) %>%
  complete(age, sex, fill = list(n = 0)) %>% 
  ggplot(aes(age, n)) + geom_col(aes(fill = sex), position = "dodge") +
    theme_classic() + 
    scale_fill_manual(values=c("#000000", "#CCCCCC"), name = "Sex") + 
    labs(x = "Age", y = "Number of observations") +   
    theme(legend.position=c(0.4,0.8),
    plot.title = element_text(size = 10),
    legend.title=element_text(size=15),
    axis.title=element_text(size=15),
    legend.key.size = unit(1.13, "cm"),
    legend.direction="vertical",
    legend.text=element_text(size=15))

p07<-data[(data$yy=="2007") & (data$item=="piercing"),]
summary(p07)

subp07<-p07 %>%  
  group_by(age) %>% 
  count(sex) %>% 
  ungroup %>% 
  mutate(age = factor(age)) %>%
  complete(age, sex, fill = list(n = 0)) %>% 
  ggplot(aes(age, n)) + geom_col(aes(fill = sex), position = "dodge") +
    theme_classic() + 
    scale_fill_manual(values=c("#000000", "#CCCCCC"), name = "Sex") + 
    labs(x = "Age", y = "Number of observations") +   
    theme(legend.position=c(0.4,0.8),
    plot.title = element_text(size = 10),
    legend.title=element_text(size=15),
    axis.title=element_text(size=15),
    legend.key.size = unit(1.13, "cm"),
    legend.direction="vertical",
    legend.text=element_text(size=15))

p15<-data[(data$yy=="2015") & (data$item=="piercing"),]

subp15<-p15 %>% 
  group_by(age) %>% 
  count(sex) %>% 
  ungroup %>% 
  mutate(age = factor(age)) %>%
  complete(age, sex, fill = list(n = 0)) %>% 
  ggplot(aes(age, n)) + geom_col(aes(fill = sex), position = "dodge") +
    theme_classic() + 
    scale_fill_manual(values=c("#000000", "#CCCCCC"), name = "Sex") + 
    labs(x = "Age", y = "Number of observations") +   
    theme(legend.position=c(0.4,0.8),
    plot.title = element_text(size = 10),
    legend.title=element_text(size=15),
    axis.title=element_text(size=15),
    legend.key.size = unit(1.13, "cm"),
    legend.direction="vertical",
    legend.text=element_text(size=15))

grid.arrange(overp, subp07, subp15)

The code I have posted gives me the following figure.

What I am trying to do is plot the frequencies for females in 2007 and 2015 and males in 2007 and 2015 on top of the barplot for total frequencies (where this is also reflected in the legend). Is there a way to do that in R using ggplot2 ?

UPDATE: I tried using the geom_smooth and geom_line functions to add the lines to my ggplot as suggested in the comments and as other solutions to users questions, but I get the following error:

Error: Discrete value supplied to continuous scale

I created a new data frame for a subset that I would like to plot:

df<-data.frame(age=c(15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,40,50,60), val=c(0,5,13,77,70,106,62,51,46,27,46,16,22,16,14,48,21, 3,4))

And then added it to the ggplot code:

overallpierc %>%
  filter(age != "15") %>% 
  group_by(age) %>% 
  count(sex) %>% 
  ungroup %>% 
  mutate(age = factor(age)) %>%
  complete(age, sex, fill = list(n = 0)) %>% 
  ggplot(aes(age, n)) +     
    geom_line(data=df,aes(x=as.numeric(age),y=val),colour="blue") +
    geom_col(aes(fill = sex), position = "dodge") +
    theme_classic() + 
    scale_fill_manual(values=c("#000000", "#CCCCCC"), name = "Sex") + 
    labs(x = "Age", y = "Number of observations") +   
    theme(legend.position=c(0.4,0.8),
    plot.title = element_text(size = 10),
    legend.title=element_text(size=15),
    axis.title=element_text(size=15),
    legend.key.size = unit(1.13, "cm"),
    legend.direction="vertical",
    legend.text=element_text(size=15))

Others have encountered similar issues and used as.numeric to solve the problem. However, age needs to be treated as a factor for the purposes of plotting.

Answer 1

Based on our discussion in the comments, let's try stacked bars and facets. I think it works but you can decide for yourself.

The stacked bar has the advantage of showing both proportions and total count in the same bar. To compare years, a facet grid places years in rows, so the eye can scan downwards to compare the same age in different years. Note that I kept age as a continuous variable here, rather than a factor.

library(dplyr)
library(ggplot2)
data30g %>% 
  count(yy, sex, age) %>% 
  ggplot(aes(age, n)) + 
    geom_col(aes(fill = sex)) + 
    facet_grid(yy ~ .) + 
    theme_bw() + 
    scale_fill_manual(values = c("#000000", "#cccccc"))

Not bad - I can see straight away, for example, an increase in both total and female count at age 30 over time, but perhaps a little small and crowded.

We can use a facet wrap instead of a grid to make the bars clearer, but at the expense of quick visual comparison across years.

data30g %>% 
  count(yy, sex, age) %>% 
  ggplot(aes(age, n)) + 
    geom_col(aes(fill = sex)) + 
    facet_wrap(~yy, ncol = 2) + 
    theme_bw() + 
    scale_fill_manual(values = c("#000000", "#cccccc"))

One more example which does not address your question in terms of total counts or barplots - but I thought it might be of interest. This code generates a "heatmap" style of plot which is poor for quantitative comparison, but can sometimes give a quick visual impression of interesting features. I think it shows, for example, that females aged 20 in 2014 have the highest total count.

data30g %>% 
  count(yy, sex, age) %>% 
  ggplot(aes(factor(age), yy)) + 
    geom_tile(aes(fill = n)) + 
    facet_grid(sex ~ .) + 
    scale_fill_gradient2() + 
    scale_y_reverse(breaks = 2006:2015) + 
    labs(x = "age", y = "Year")

EDIT:

Based on further discussions in the comments, here is one way to plot age as a factor, using bars for sexes, overlaid with a line for the totals and split by year.

overallpierc %>% 
  count(yy, sex, age) %>% 
  ggplot() + 
    geom_col(aes(factor(age), n, fill = sex), position = "dodge") +
    stat_summary(aes(factor(age), n), fun.y = "sum", geom = "line", group = 1) + 
  facet_grid(yy ~ .)