Bar chart from two data frames in R

Question

My aim is to do a basic bar chart where the data comes from two data frames that have the same variables. The plot should look like the one below, but without the two shortcomings. There should be legends for the data frames, and the error bars should be in the middle of the bars. Do you know how to make it? Below is the code that produced that plot, and I don't think bind_rows is the optimal solution here.

The code ↓

bind_rows(B, C, .id = "id") %>%
  filter(question %in% c("Q1", "Q2")) %>%
  ggplot(aes(x = question, y = mean)) +
  geom_bar(aes(fill = id), stat = "identity", position = "dodge2", width = 0.5) +
  geom_errorbar(aes(ymin = mean - sd, ymax = mean + sd, width = 0.1, colour = id)) +
  coord_flip()

The data frames ↓

structure(list(question = c("Q1", "Q10", "Q11", "Q12", "Q2", 
"Q3", "Q4", "Q5", "Q6", "Q7", "Q8", "Q9"), n = c(204L, 204L, 
204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L), 
    mean = c(5.22549019607843, NA, 4.95098039215686, 4.39705882352941, 
    5.47058823529412, 5.51470588235294, 4.50490196078431, 4.92647058823529, 
    4.40686274509804, 5.56862745098039, 5.56372549019608, 5.23529411764706
    ), sd = c(1.1524816893289, NA, 1.31214449357814, 1.5422430010719, 
    1.12039650223724, 1.15104553532809, 1.37714471881058, 1.34621721218454, 
    1.30030385262334, 0.871099231072865, 0.830963499839951, 1.36945187401243
    )), row.names = c(NA, 12L), class = c("tbl_df", "tbl", "data.frame"
))

structure(list(question = c("Q1", "Q10", "Q11", "Q12", "Q2", 
"Q3", "Q4", "Q5", "Q6", "Q7", "Q8", "Q9"), n = c(13L, 13L, 13L, 
13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L), mean = c(5.38461538461539, 
4.38461538461539, 4.69230769230769, 4.30769230769231, 5.15384615384615, 
5.38461538461539, 4.76923076923077, 5.30769230769231, 4.53846153846154, 
5.61538461538461, 5.69230769230769, 4.92307692307692), sd = c(1.26085034391223, 
1.44559454541846, 1.03155347127648, 1.60128153805087, 0.898717034272917, 
1.12089707663561, 1.01273936708367, 0.85485041426511, 0.967417922046845, 
1.26085034391223, 0.85485041426511, 1.84668795692624)), row.names = c(NA, 
12L), class = c("tbl_df", "tbl", "data.frame"))

Answer 1

This post is tagged along with ggplot2 and tidyverse. However, if you want to just use the base package, here's one approach (a bit lengthy and tricky, but it does the work just fine):

df1<-structure(list(question = c("Q1", "Q10", "Q11", "Q12", "Q2", "Q3", "Q4", "Q5", "Q6", "Q7", "Q8", "Q9"),
                n = c(204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L, 204L),
                mean = c(5.22549019607843, NA, 4.95098039215686, 4.39705882352941, 5.47058823529412,
                5.51470588235294, 4.50490196078431, 4.92647058823529, 4.40686274509804, 5.56862745098039,
                5.56372549019608, 5.23529411764706), sd = c(1.1524816893289, NA, 1.31214449357814, 1.5422430010719,
                1.12039650223724, 1.15104553532809, 1.37714471881058, 1.34621721218454, 1.30030385262334,
                0.871099231072865, 0.830963499839951, 1.36945187401243)), row.names = c(NA, 12L), 
                class = c("tbl_df", "tbl", "data.frame"))
df2<-structure(list(question = c("Q1", "Q10", "Q11", "Q12", "Q2", "Q3", "Q4", "Q5", "Q6", "Q7", "Q8", "Q9"),
                n = c(13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L, 13L),
                mean = c(5.38461538461539, 4.38461538461539, 4.69230769230769, 4.30769230769231, 5.15384615384615,
                5.38461538461539, 4.76923076923077, 5.30769230769231, 4.53846153846154, 5.61538461538461, 
                5.69230769230769, 4.92307692307692), sd = c(1.26085034391223, 1.44559454541846, 1.03155347127648,
                1.60128153805087, 0.898717034272917, 1.12089707663561, 1.01273936708367, 0.85485041426511,
                0.967417922046845, 1.26085034391223, 0.85485041426511, 1.84668795692624)), row.names = c(NA, 12L),
                class = c("tbl_df", "tbl", "data.frame"))

First, we merge the two dataframes:

df3<-merge(df1,df2,by="question")

#df3
#question n.x   mean.x      sd.x n.y   mean.y      sd.y
#1        Q1 204 5.225490 1.1524817  13 5.384615 1.2608503
#2       Q10 204       NA        NA  13 4.384615 1.4455945
#3       Q11 204 4.950980 1.3121445  13 4.692308 1.0315535
#4       Q12 204 4.397059 1.5422430  13 4.307692 1.6012815
#5        Q2 204 5.470588 1.1203965  13 5.153846 0.8987170
#6        Q3 204 5.514706 1.1510455  13 5.384615 1.1208971
#7        Q4 204 4.504902 1.3771447  13 4.769231 1.0127394
#8        Q5 204 4.926471 1.3462172  13 5.307692 0.8548504
#9        Q6 204 4.406863 1.3003039  13 4.538462 0.9674179
#10       Q7 204 5.568627 0.8710992  13 5.615385 1.2608503
#11       Q8 204 5.563725 0.8309635  13 5.692308 0.8548504
#12       Q9 204 5.235294 1.3694519  13 4.923077 1.8466880

Then, we store our data in the form of a matrix:

mat<-as.matrix(cbind(df3$mean.x,df3$mean.y))

#mat
#         [,1]     [,2]
# [1,] 5.225490 5.384615
# [2,]       NA 4.384615
# [3,] 4.950980 4.692308
# [4,] 4.397059 4.307692
# [5,] 5.470588 5.153846
# [6,] 5.514706 5.384615
# [7,] 4.504902 4.769231
# [8,] 4.926471 5.307692
# [9,] 4.406863 4.538462
#[10,] 5.568627 5.615385
#[11,] 5.563725 5.692308
#[12,] 5.235294 4.923077

Finally, we plot the barplots:

#par(mar=c(5,4,4,5)+0.1)
mid<-barplot(t(mat),beside=T,names.arg=df3$question,cex.names=0.8,
            horiz=T,xlim=c(0,7),col=c('#44c1f2','#f28744'),
            xlab='mean',ylab='question')

Now, for plotting error bars in the barplot from the base package, we can do the following (taking inspiration from this post by @Laryx Decidua):

arrows(x0=df1$mean-df1$sd, y0=mid[1,], x1=df1$mean+df1$sd, y1=mid[1,],
        code=3, angle=90, length=0.04, col="red")
arrows(x0=df2$mean-df2$sd, y0=mid[2,], x1=df2$mean+df2$sd, y1=mid[2,],
        code=3, angle=90, length=0.04, col="blue")

Now add the legend:

legend(7,21,c('1','2'),fill=c('#44c1f2','#f28744'),xpd=T,title='ID')

Which brings us to a plot like this:

Answer 2

I think dplyr::bind_rows() works perfectly fine here. To align the grouped bars and the error bars, use position_dodge() for both layers.

bind_rows(B, C, .id = "id") %>%
  filter(question %in% c("Q1", "Q2")) %>% 
  ggplot() +
  aes(mean, question, fill = id, xmin = mean - sd, xmax = mean + sd) +
  geom_col(position = "dodge2", width = 0.5) +
  geom_errorbar(position = position_dodge2(padding = 0.5), width = 0.5) +
  theme_minimal()

position_dodge2() is also needed if you want to add additional layers.

last_plot() +
  geom_point(position = position_dodge2(width = 0.5))