How to insert text label on faceted time series data using the forecast package in R?-CodePudding

I am doing time series analysis in R using the forecast package. The forecast::autoplot() creates a ggplot object.

I have two questions about plotting decomposed series:

How do I place text in the trend panel (next to the vertical red line). Using ggplot or other packages?
How can I rename the "data" panel to be capitalized as "Data"?

Here is my code:

library(forecast)    

# Load in AirPassengers time series
data("AirPassengers")

# Decomopose time series
decomposed_ts <- decompose(AirPassengers)

# Plot decomposed time series
forecast::autoplot(decomposed_ts)   geom_vline(xintercept = 1954, colour = "red")

For Q1, I tried to use annotate() but it places the same text labels on each facet, whereas I only want one label on only the trend panel (see screenshot). I also tried to use the ggtext package, but could not get it to work. I couldn't figure out what the faceted variable is because forecast::autoplot() automatically created the faceting.

For Q2. I saved the plot as a variable i.e. myPlot <- forecast::autoplot(decomposed_ts) geom_vline(xintercept = 1954, colour = "red"). I then tried myPlot$ looking for where the "data" facet is named.

CodePudding user response：

I'm sure someone smarter than me could dig into the architecture of a ggplot2 plot, but you could make your own autoplot function to keep it a little simpler(ish). Here I added a facet label argument to solve your problem:

library(tidyverse)
library(forecast)    


my_new_autoplot <- function(decomp, 
                            facet_labels = c("data", "trend", "seasonal", "remainder")){
  list(
tibble(Time = time(decomp$x),
       val = decomp$x,
       type = facet_labels[1]),

tibble(Time = time(decomp$trend),
       val = decomp$trend,
       type = facet_labels[2]),

tibble(Time = time(decomp$seasonal),
       val = decomp$seasonal,
       type = facet_labels[3]),

tibble(Time = time(decomp$random),
       val = decomp$random,
       type = facet_labels[4]) |>
  mutate(t1 = pmax(0, val),
         t2 = pmin(val, 0),
         val = 0)
) |>
  bind_rows() |>
  mutate(type = factor(type, levels = facet_labels)) |>
  ggplot(aes(Time, val)) 
  geom_path() 
  geom_errorbar(aes(ymin = t2, ymax = t1)) 
  facet_wrap(~type, ncol = 1, scales = "free_y", strip.position = "left") 
  labs(y = "")
}

my_new_autoplot(decomposed_ts)   
  geom_vline(xintercept = 1954, colour = "red")

my_new_autoplot(decomposed_ts, 
                facet_labels = c("my_data!", "trend..", "~seasonal~", "rEmainDeR"))   
  geom_vline(xintercept = 1954, colour = "red")

And if we wanted to add text, we can do that too:

my_plot <- my_new_autoplot(decomposed_ts, 
                facet_labels = c("my_data!", "trend..", "~seasonal~", "rEmainDeR"))   
  geom_vline(xintercept = 1954, colour = "red")


add_text <- function(plot, facet, text, x, y, ...){
  text_data <- tibble(Time = x, val = y, type = facet, label = text) |>
    mutate(type = factor(type, levels = levels(plot$data$type)))
  
  plot 
    geom_text(data = text_data, aes(label = label), ...)
}

add_text(plot = my_plot, 
         facet = "rEmainDeR", 
         text = "This is red", 
         x = 1955, 
         y = 40,
         color = "red")

CodePudding user response：

You can add geom_text to the original auto_plot:

annotation <- data.frame(
  x = c(1950),
  y = c(250),
  parts = c("trend"),
  label = c("label 1")
)

forecast::autoplot(decomposed_ts)  
  geom_vline(xintercept = 1954, colour = "red")  
  geom_text(data = annotation, aes(x = x, y = y, label = label))