How to replicate this plot with R?

Question

I need to make several time-series graphs in the same format as shown in Fig. I'm interested in the format itself.

I will be grateful for your help.

Here a minimal example with simulated data

    set.seed(1234)
    nobs<-50
    x<-rnorm(nobs,0,10)
    t<-seq(1,nobs)
    data<-ts(20+x+t, freq=4, start=c(2004,1))
    par(family="serif")
    plot(data, lwd=2, col="blue4", ylab="Millions")
    title(main="Gráfico 1. Evolución del ...")
    legend(2010,10,"GDP", col="blue4",lwd=2)
    abline(lm(data~t), lwd=2, col="red") # This does not work

Mainly, the same labels on the x-axis and y-axis, and the same legend position. — Héctor Garrido, Sep 19 '16 at 00:38
Well, and the regression line. I tried something similar with the abline function, but it did not work. — Héctor Garrido, Sep 19 '16 at 00:40
I think it would still be useful for you to provide a minimal example. This helps other users to help you with exactly the parts of the plot that you are having difficulty with, given your existing code (e.g. if you are using `ggplot` or `base` graphics). — Weihuang Wong, Sep 19 '16 at 00:51
Is your data actually a time series object in R, or does it start out as, say, as csv file that you load into an R data frame and then convert to a time series object in R? — eipi10, Sep 19 '16 at 02:52

score 6 · Accepted Answer · answered Sep 19 '16 at 06:08

I assume that your data is a ts object, as in your example. Here is an approach using base graphics. We start off with a minimal plot, then layer on elements. First, the minimal plot, making sure to leave lots of space at side 1 (i.e. bottom):

par(mar = c(8, 4, 4, 2), family = "serif", las = 1)
plot(data, 
  frame = FALSE,
  xaxt = "n", 
  ylab = "Millones de pesos", xlab = "", 
  main = "Gráfico 1. Evolución del ...",
  col = "darkblue", lwd = 2)
fit <- forecast::tslm(data ~ trend)
lines(fitted(fit), col = "red", lwd = 2)

Note that above, we use forecast::tslm to get the time trend for the time series. Now, add the x-axis elements. I use mtext to insert the tick labels:

axis(1, at = seq(2003.875, 2015.875, 1), tck = -.2, labels = FALSE)
mtext(rep(c("ener-mar", "abril-jun", "jul-sep", "oct-dic"), l = length(data)),
  side = 1, line = 0.25, las = 2,
  at = seq(tsp(data)[1], tsp(data)[2], by = 1/tsp(data)[3]),
  cex = .6)
mtext(2004:2016, side = 1, line = 3, at = seq(2004.3, 2016.3), cex = .7)
title(xlab = "Trimestres", line = 5, cex.lab = .8)

Finish off with the legend:

par(xpd = TRUE)
legend(2010, -27, c("PIB", "Tendencia PIB"), 
  bty = "n",
  xjust = .5,
  lty = c(1, 1),
  col = c("darkblue", "red"),
  cex = .7,
  horiz = TRUE)

Output:

score 3 · Answer 2 · edited May 23 '17 at 12:00

Here's a ggplot2 approach that comes close to the plot format in your question:

library(ggplot2)

# Fake data
set.seed(1234)
nobs<-52
t<-seq(1,nobs)
x<-rnorm(nobs,0,10) + t + 20
dat = data.frame(year=rep(2004:2016, each=4), qtr=rep(1:4, 13), x = x)

ggplot(dat, aes(paste(year,qtr), x)) +
  geom_vline(xintercept=seq(4.5,100,4), colour="grey80",lwd=0.3) +
  geom_line(aes(group=1)) +
  annotate(seq(2.5, 2.5 + 12*4, length.out=13), -2, label=2004:2016, 
           geom="text", colour="grey30", size=3.5) +
  theme_bw(base_size=10) +
  theme(axis.text.x=element_text(angl=-90, vjust=0.5, colour="grey30"),
        panel.grid.major.x=element_blank()) +
  scale_x_discrete(name="Quarter",
                   labels=rep(paste0(month.abb[seq(1,12,3)],"-",month.abb[seq(3,12,3)]),20)) +
  coord_cartesian(xlim=c(0,13*4 + 1), ylim=c(-4,max(dat$x + 2)), expand=FALSE)

Facetting comes closer to the formatting you wanted, but there's no way (AFAIK) within the normal ggplot workflow to connect lines across the facets. In the plot below, the vertical lines show the borders between each facet. There are (somewhat complicated) ways to try and connect lines across facets, but I didn't make it that far in the example below.

dat$qtr = rep(paste0(month.abb[seq(1,12,3)],"-",month.abb[seq(3,12,3)]),13)
dat$qtr = factor(dat$qtr, levels=paste0(month.abb[seq(1,12,3)],"-",month.abb[seq(3,12,3)]))

p=ggplot(dat, aes(qtr, x)) +
  geom_line(aes(group=1)) +
  coord_cartesian(xlim=c(0.5,4.5), expand=FALSE, ylim=c(-2,80)) +
  facet_grid(. ~ year, switch="x") +
  annotate(x=c(0.5, 4.5), xend=c(0.5,4.5), y=-16, yend=-2, geom="segment", 
           colour="grey70", size=0.3) +
  theme_bw(base_size=10) +
  theme(panel.margin=unit(0,"lines"),
        axis.text.x=element_text(angl=-90, vjust=0.5, colour="grey30"),
        panel.border=element_rect(colour="grey70", size=0.3),
        panel.grid.major=element_blank(),
        strip.background=element_rect(fill="grey90", colour="grey20"),
        axis.line=element_line(colour="black")) +
  labs(x="Quarter")

# Turn off clipping and draw plot
gt <- ggplot_gtable(ggplot_build(p))
gt$layout$clip[gt$layout$name=="panel"] <- "off"
grid.draw(gt)

How to replicate this plot with R?

2 Answers2