**"R-bloggers" via Tal Galili in Google Reader**, and kindly contributed to R-bloggers)

On the `ggplot2` mailing-list the following question was asked:

How to create a back-to-back bar chart with ggplot2?

For anyone who don’t know what I am talking about, have a look on a recent paper from the EU. I’d like to create plots like the graphs 5,6,18 in the paper.

An example graph from the above report is below:

Let’s create the same graph in `ggplot2`.

I was not able to find the exact dataset used to plot the graph above, and used instead the Eurostat “EU27 trade by BEC product group since 1999″ dataset which has a very similar data structure.

Access the subset used in this post in here.

> library(ggplot2) |

> trade <- read.csv("trade.csv", header = TRUE, stringsAsFactors = FALSE) |

Calculate monthly total trade balances.

> balance <- ddply(trade, .(Time), summarise, balance = sum(EXP - IMP)) |

Convert data from wide format to long format for plotting.

> trade.m <- melt(trade, id.vars = c("BEC", "Time")) |

> ggplot(trade.m, aes(Time)) + geom_bar(subset = .(variable == "EXP"), aes(y = value, fill = BEC), stat = "identity") + geom_bar(subset = .(variable == "IMP"), aes(y = -value, fill = BEC), stat = "identity") + xlab("") + scale_y_continuous("Export - Import", formatter = "comma") |

Add horizontal line at 0 for easier reading, plot the trade balance line and make the x-axis labels more readable.

> labels <- gsub("20([0-9]{2})M([0-9]{2})", "\2n\1", trade.m$Time) > last_plot() + geom_line(data = balance, aes(Time, balance, group = 1), size = 1) + geom_hline(yintercept = 0, colour = "grey90") + scale_x_discrete(labels = labels) |

The legend explanation (too large to fit on the plot):

- CAP Capital goods
- CNS Consumption goods
- CTR Consumption goods plus motor spirit and passenger motor cars
- INT Intermediate goods

There shouldn’t be any negative numbers on the y-axis, we need a custom formatter to convert these to positive.

> commapos <- function(x, ...) { format(abs(x), big.mark = ",", trim = TRUE, scientific = FALSE, ...) } |

> last_plot() + scale_y_continuous(formatter = "commapos") |

I have to agree with some of the posters on the mailing list, that this type of chart is not very easy to follow – it is impossible to compare individual values. Another useful way of showing information would be to use facets:

> ggplot(trade.m, aes(Time)) + geom_bar(aes(y = value), stat = "identity") + facet_grid(variable ~ BEC) + scale_x_discrete("", labels = labels) + scale_y_continuous(formatter = "comma") |

Or combine export and import values on one plot showing the comparative trend between the different goods categories.

> ggplot(trade.m, aes(Time, colour = variable, group = variable)) + geom_line(aes(y = value)) + facet_grid(BEC ~ .) + scale_x_discrete("", labels = labels) + scale_y_continuous(formatter = "comma") |

The same graph with varying scales across panels.

> last_plot() + facet_grid(BEC ~ ., scales = "free_y") |

** Update:** As suggested in the comments it would be good to have total values added to the last plot.

In order to do so the totals are first calculated and then appended to the original dataframe.

> total <- ddply(trade.m, .(Time, variable), summarise, value = sum(value), BEC = "Total") > total <- rbind(trade.m, total) |

Plotting the new dataframe is just a question of changing the source of the last plot.

> last_plot() %+% total |

**leave a comment**for the author, please follow the link and comment on his blog:

**"R-bloggers" via Tal Galili in Google Reader**.

R-bloggers.com offers

**daily e-mail updates**about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...