How can I change the fabletools autoplot or autolayer level legend title?

General
1

I am trying to change the title of the legend for the confidence intervals produced by the fable autoplot or autolayer functions to say "Confidence Interval" instead of "level".

Here's a reproducible example:

fit_consBest <- us_change |>
  model(
    lm = TSLM(Consumption ~ Income + Savings + Unemployment)
  )

Increase = new_data(us_change, 4) |>
    mutate(Income=1, Savings=0.5, Unemployment=0)

fc <- forecast(fit_consBest, new_data = Increase)

us_change |>
  autoplot(Consumption) +
  autolayer(fc) +
  labs(title = "US consumption", 
       y = "% change",
       level = "Confidence Interval")

image
image891×382 3.94 KB

I tried adding level = "Confidence Interval" to the labs function based on the answer to this question: Change legend title of forecast plot - #3 by budugulo but that was using the fpp3 library and it doesn't seem to work with fable or fabletools. I have tried reading google search results and asking ChatGPT back and forth for hours and I can't figure out how to change that title.

Thanks so much for your help!

2

First, some corrections.

  1. These are not confidence intervals. They are prediction intervals. There is a difference.
  2. The fpp3 package loads the fable and fabletools packages (and a few others). Any code that loads fpp3 will also be using the fable package and its friends.

Now, to answer the question. The autoplot() function is returning a ggplot object, so you can use ggplot functions to modify it. In this case, you need to replace the title of the fill_ramp guide.

library(fpp3)
#> ── Attaching packages ──────────────────────────────────────────── fpp3 1.0.1 ──
#> ✔ tibble      3.2.1     ✔ tsibble     1.1.5
#> ✔ dplyr       1.1.4     ✔ tsibbledata 0.4.1
#> ✔ tidyr       1.3.1     ✔ feasts      0.4.1
#> ✔ lubridate   1.9.3     ✔ fable       0.4.0
#> ✔ ggplot2     3.5.1
#> ── Conflicts ───────────────────────────────────────────────── fpp3_conflicts ──
#> ✖ lubridate::date()    masks base::date()
#> ✖ dplyr::filter()      masks stats::filter()
#> ✖ tsibble::intersect() masks base::intersect()
#> ✖ tsibble::interval()  masks lubridate::interval()
#> ✖ dplyr::lag()         masks stats::lag()
#> ✖ tsibble::setdiff()   masks base::setdiff()
#> ✖ tsibble::union()     masks base::union()

fit_consBest <- us_change |>
  model(
    lm = TSLM(Consumption ~ Income + Savings + Unemployment)
  )

Increase <- new_data(us_change, 4) |>
  mutate(Income = 1, Savings = 0.5, Unemployment = 0)

fc <- forecast(fit_consBest, new_data = Increase)

us_change |>
  autoplot(Consumption) +
  autolayer(fc) +
  labs(title = "US consumption", y = "% change") +
  guides(fill_ramp = guide_legend(title = "Prediction Interval"))

Your code could be be slightly simpler by using autoplot() on the forecast object, rather than on the data. Like this:

fc |>
  autoplot(us_change) +
  labs(title = "US consumption", y = "% change") +
  guides(fill_ramp = guide_legend(title = "Prediction Interval"))

Created on 2024-10-22 with reprex v2.1.1

.

2 Likes
3

THANK YOU! I seriously can't thank you enough. I spent hours trying to figure this out yesterday and even though I knew it could be edited with ggplot, I couldn't seem to figure out the right combination of ggplot functions and variable names... Maybe I need to do a deep study of ggplot.

Here's my actual code. I couldn't easily share the data so I just made a simple reprex to explain what I was trying to do. The last line was added based on your answer and fixed my problem.

cuny_ts |>
  autoplot(enrl, color = "#0033A1", linewidth = 1) +
  geom_line(data = fitted(fit_ensemble),
            aes(y = .fitted), alpha = 1,  color = "#FFB71B", linewidth = 1) + 
  autolayer(fc_ensemble, alpha = 0.35, level = 80,  color = "#FFB71B", linewidth = 1) + # makes the prediction interval transparent
  autolayer(fc_ensemble, alpha = 1, level = NULL,color = "#FFB71B", linewidth = 1) +  # makes the forecast line the correct solid CUNY branded color not a transparent version of it
  labs(x = "Year",
       y = "Number of Students",
       title = "CUNY Total Undergraduate Enrollment from 1990-2023 With 10-Year Forecast") +
  scale_y_continuous(labels = scales::comma_format()) +
  theme_bw(base_size = 10)  +
  guides(fill_ramp = guide_legend(title = "Prediction Interval"))

Unfortunately I am not allowed to share the plot that it produces publicly.

And THANK YOU again :pray: for correcting my understanding of the prediction interval vs confidence interval. I really appreciate that and will be correcting my report. I read Forecasting: Principles and Practice (3rd ed) almost cover to cover while working on this project to try to make sure I did everything correctly but I still feel a little shaky on some things and know I have a lot more to learn...