Fable: Extracting the p,d,q specification from an ARIMA model

Question

I've been using the tidy forecasting package fable (which has been so useful).

I was wondering if there was an easy way to extract that the p,d,q values from a mable.

Using the data on this guide as an example https://www.mitchelloharawild.com/blog/fable/

library(tidyverse)
library(tsibble)
library(fable)

tourism_state <- tourism %>% 
  group_by(State) %>% 
  summarise(Trips = sum(Trips))

fit <- tourism_state %>% 
  model(arima = ARIMA(Trips))

> fit
# A mable: 8 x 2
# Key:     State [8]
  State                                 arima
  <chr>                               <model>
1 ACT                          <ARIMA(0,1,1)>
2 New South Wales    <ARIMA(0,1,1)(0,1,1)[4]>
3 Northern Territory <ARIMA(1,0,1)(0,1,1)[4]>
4 Queensland                   <ARIMA(2,1,2)>
5 South Australia    <ARIMA(1,0,1)(0,1,1)[4]>
6 Tasmania           <ARIMA(0,0,3)(2,1,0)[4]>
7 Victoria           <ARIMA(0,1,1)(0,1,1)[4]>
8 Western Australia            <ARIMA(0,1,3)>

I know the specifications are stored under model[[1]]$fit$spec but I can't figure out a way to extract them if I have a large list of models

Ideally I'd like

  State                                 arima       p     d       q
  <chr>                               <model>
1 ACT                          <ARIMA(0,1,1)>       0     1       1
2 New South Wales    <ARIMA(0,1,1)(0,1,1)[4]>       0     1       1
3 Northern Territory <ARIMA(1,0,1)(0,1,1)[4]>       1     0       1
4 Queensland                   <ARIMA(2,1,2)>       
5 South Australia    <ARIMA(1,0,1)(0,1,1)[4]>       and so on....
6 Tasmania           <ARIMA(0,0,3)(2,1,0)[4]>
7 Victoria           <ARIMA(0,1,1)(0,1,1)[4]>
8 Western Australia            <ARIMA(0,1,3)>

Thanks!

Majordomo

Answer 1

What about this?

# specifically needed libraries from tidyverse
library(dplyr)
library(purrr)

fit %>%
  mutate(map_dfr(arima, c("fit", "spec")))

#> # A mable: 8 x 10
#> # Key:     State [8]
#>   State                                 arima     p     d     q     P     D     Q constant period
#>   <chr>                               <model> <int> <int> <int> <int> <int> <int> <lgl>     <dbl>
#> 1 ACT                          <ARIMA(0,1,1)>     0     1     1     0     0     0 FALSE         4
#> 2 New South Wales    <ARIMA(0,1,1)(0,1,1)[4]>     0     1     1     0     1     1 FALSE         4
#> 3 Northern Territory <ARIMA(1,0,1)(0,1,1)[4]>     1     0     1     0     1     1 FALSE         4
#> 4 Queensland                   <ARIMA(2,1,2)>     2     1     2     0     0     0 FALSE         4
#> 5 South Australia    <ARIMA(1,0,1)(0,1,1)[4]>     1     0     1     0     1     1 FALSE         4
#> 6 Tasmania           <ARIMA(0,0,3)(2,1,0)[4]>     0     0     3     2     1     0 FALSE         4
#> 7 Victoria           <ARIMA(0,1,1)(0,1,1)[4]>     0     1     1     0     1     1 FALSE         4
#> 8 Western Australia            <ARIMA(0,1,3)>     0     1     3     0     0     0 FALSE         4

It works on R >= 4.0 and dplyr >= 1.0.

The arima column is a list. We can use map to extract data from lists.

map will return a list itself, but with map_dfr you can return a dataframe which mutate will interprete as a new set of columns to add to the original dataframe.

Note that with this code the output and the input keep the same class (mable).

---

UPDATE

Since tidyverse changes every 5 days, if you have troubles with this solution (since map_dfr is currently superseded), try with this:

fit %>%
  mutate(map(arima, c("fit", "spec")) |> list_rbind())

It should work the same.

Cupidity