Multiple x-axis labels for time-series data

Question

I am able to plot the time-series data using ggplot2. However, I want to highlight the seasonal information alongwith the time-series data.

Here's my code:

library(zoo)
library(ggplot2)

a <- read.table(text = "
       Season Quarter  Sales
       Season1  2014Q1  20 
       Season1  2014Q2  40 
       Season1  2014Q3  60 
       Season1  2014Q4  80 
       Season2  2015Q1  30 
       Season2  2015Q2  40 
       Season2  2015Q3  80 
       Season3  2015Q4  90 
       Season3  2016Q1  100 
       Season3  2016Q2  120 
       Season3  2016Q3  140
     ", header = TRUE, sep = "")
a$Quarter<-as.yearqtr(a$Quarter)
a$Quarter<-as.Date(a$Quarter)

ggplot(data=a,aes(x=Quarter, y=Sales)) +
       geom_line()

This works well in that I am able to draw a time-series data.

Now, I want to label what constitutes Season 1, 2 etc. One way to do this would be to use color or linetype. However, this doesn't seem to work because it breaks the continuity of the time-series.

# doesn't work...
ggplot(data=a,aes(x=Quarter, y=Sales)) +
       geom_line(aes(linetype=Season))

On the other hand, I like how Excel plots this graph in just two clicks. It creates a beautiful graph that shows seasonal information on x-axis along with dates. It essentially creates a 3-layered x-axis.

I have two questions on this topic:

Question 1: Using ggplot, how can I use linetype (or color) in ggplot to create continuous graph (i.e. without breaks)? I'd prefer linetype over color. As an example and to answer the comment: here's the graph I created using a different set of data.

df <- data.frame(x = 1:3, y = 1:3, z = c(1,3,5))
ggplot(df, aes(x, y, color = factor(z))) +
       geom_line(aes(group = 1))

I am unable to replicate above behavior for time-series data. Here's the graph I got from above code:

Question 2: Using ggplot, how can I create a multi-level x-axis (similar to what Excel did for me) that shows Seasonal information with dates? {Please see Excel graph that I created.} I do NOT want to create a legend using this option. I also want to clarify that I'd appreciate if we don't use hacking methods by applying annotate (or possibly geom_text) methods to put these multi-level labels by adjusting and re-adjusting x- and y- values to fit them. This is because it defeats the purpose of using programming language to draw the graph, and it won't work as the data change.

If you have any questions, please let me know. I'd appreciate your thoughts. I am an absolute beginner with ggplot2. It's been only 5 days since I have transitioned from Excel and STATA to ggplot. So, I apologize if my question is too basic.

I researched this topic on SO and couldn't anything that is close enough. For instance, this thread talks about changing ticks, but not what I am looking for.

While your intent is clear, in particular for your `linetype` example, `ggplot2` is doing the right thing -- for example, what `linetype` would you expect the segment connecting seasons 1 & 2 to have? — tchakravarty, Aug 29 '16 at 07:09
@tchakravarty Thanks for your feedback. Now, I am thinking that if we can show the continuity by using colors, it would help. I will add clarification above. — watchtower, Aug 29 '16 at 07:14
The same logic applied to colors as well. I think that your best bet is to try and replicate the hierarchical axes that Excel is creating for you. There are some previous solutions on SO [here](http://stackoverflow.com/q/18165863/1414455) and [here](http://stackoverflow.com/q/5704725/1414455). — tchakravarty, Aug 29 '16 at 07:16
@tchakravarty I have added the graph. I am unable to replicate this for time-series data. I am really lost... — watchtower, Aug 29 '16 at 07:22
@ss0208535 Please ask one question per SO question. Your Q2 is a duplicate of [this](http://stackoverflow.com/questions/20571306/multi-row-x-axis-labels-in-ggplot-line-chart). — Henrik, Aug 29 '16 at 07:51
@henrik - 2 things: **1** I don't think my question is duplicate. I've mentioned that I'm not looking to use annotate because it takes away the programming ability and makes us depend on hard-coding values. As you might see in that post, the poster has created an equation for x and hard-coded value of y, which will break the moment I use the code for actual data. **2** I thought of grouping two questions together because they are very well related. In the past, I've been criticized for posting two different questions on the *same* dataset. I just don't want to be flagged unnecessarily. — watchtower, Aug 29 '16 at 08:19

Andrie · Accepted Answer · 2016-08-29T08:39:37.817

You can quite easily recreate the intent of your Excel plot like this:

library(plyr)
ss <- ddply(a, .(Season), summarize, date = min(Quarter))
ss$date <- as.numeric(ss$date)

ggplot(data=a,aes(x=Quarter,y=Sales)) +
  geom_line() +
  geom_vline(data = ss, aes(xintercept = date), colour = "grey50") +
  geom_text(data = ss, aes(x = as.Date(date), y = Inf, label = Season), 
            hjust = -0.1, vjust = 1.1)

One workaround for the break in the line when using colours is to plot a continuous grey line in addition to the colour lines:

ggplot(data=a,aes(x=Quarter,y=Sales)) +
  geom_line(colour = "grey80") +
  geom_line(aes(colour = Season)) +
  geom_vline(data = ss, aes(xintercept = date), colour = "grey50") +
  geom_text(data = ss, aes(x = as.Date(date), y = Inf, label = Season), 
            hjust = -0.1, vjust = 1.1)

rcs · Answer 2 · 2016-08-29T10:37:15.370

A workaround is to modify the data frame, i.e., to add additional lines to the data frame, when there is a change in the Season column. Like this way:

library("plyr")

# add additional lines at end of every season 
tmp <- ddply(a, "Season",
             function(x) {
               x[nrow(x)+1, "Season"] <- x[nrow(x), "Season"]
               x
             })
# fill NA values with values of next season
tmp$Quarter <- na.locf(tmp$Quarter, fromLast=TRUE, na.rm=FALSE)
tmp$Sales <- na.locf(tmp$Sales, fromLast=TRUE, na.rm=FALSE)
tmp <- na.omit(tmp)   # drop last line
tmp
#     Season    Quarter Sales
# 1  Season1 2014-01-01    20
# 2  Season1 2014-04-01    40
# 3  Season1 2014-07-01    60
# 4  Season1 2014-10-01    80
# 5  Season1 2015-01-01    30
# 6  Season2 2015-01-01    30
# 7  Season2 2015-04-01    40
# 8  Season2 2015-07-01    80
# 9  Season2 2015-10-01    90
# 10 Season3 2015-10-01    90
# 11 Season3 2016-01-01   100
# 12 Season3 2016-04-01   120
# 13 Season3 2016-07-01   140

ggplot(data=tmp, aes(x=Quarter, y=Sales)) +
       geom_line(aes(colour=Season, linetype=Season))

Multiple x-axis labels for time-series data

2 Answers2