Exercise 2.1

Explore the following four time series: Bricks from aus_production, Lynx from pelt, Close from gafa_stock, and Demand from vic_elec.

Dataset Information & Time Intervals

Inspecting the help documentation (?aus_production, ?pelt, ?gafa_stock, ?vic_elec) and tsibble intervals gives the sampling frequency and units for each series:

data.frame(
  Series = c("Bricks", "Lynx", "Close", "Demand"),
  Dataset = c("aus_production", "pelt", "gafa_stock", "vic_elec"),
  Interval = c(
    format(interval(aus_production)),
    format(interval(pelt)),
    "Daily trading days (irregular)",
    format(interval(vic_elec))
  ),
  Description = c(
    "Quarterly Australian clay brick production (millions)",
    "Annual Hudson Bay Company Canadian lynx pelt records (1845-1935)",
    "Daily closing stock prices ($USD) for AAPL, AMZN, FB, GOOG (2014-2018)",
    "Half-hourly operational electricity demand (MW) in Victoria, Australia"
  )
) |>
  kable()
Series Dataset Interval Description
Bricks aus_production 1Q Quarterly Australian clay brick production (millions)
Lynx pelt 1Y Annual Hudson Bay Company Canadian lynx pelt records (1845-1935)
Close gafa_stock Daily trading days (irregular) Daily closing stock prices ($USD) for AAPL, AMZN, FB, GOOG (2014-2018)
Demand vic_elec 30m Half-hourly operational electricity demand (MW) in Victoria, Australia

Time Plots

aus_production |>
  autoplot(Bricks, color = primary_col) +
  labs(title = "Australian Clay Brick Production", x = "Quarter", y = "Millions of Bricks")

pelt |>
  autoplot(Lynx, color = primary_col) +
  scale_y_continuous(labels = comma) +
  labs(title = "Canadian Lynx Pelts Traded", x = "Year", y = "Number of Pelts")

gafa_stock |>
  autoplot(Close) +
  scale_y_continuous(labels = dollar) +
  labs(title = "GAFA Daily Closing Stock Prices (2014-2018)", x = "Date", y = "Closing Price ($USD)")

vic_elec |>
  autoplot(Demand, color = primary_col, alpha = 0.75, linewidth = 0.3) +
  scale_y_continuous(labels = comma) +
  labs(
    title = "Half-Hourly Electricity Demand in Victoria, Australia (2012-2014)",
    x = "Date and Time (30-Minute Intervals)",
    y = "Electricity Demand (MW)"
  )


Exercise 2.2

Use filter() to find what days corresponded to the peak closing price for each of the four stocks in gafa_stock.

gafa_stock |>
  group_by(Symbol) |>
  filter(Close == max(Close)) |>
  ungroup() |>
  select(Symbol, Date, Close, Volume) |>
  arrange(desc(Close)) |>
  kable(
    digits = 2,
    col.names = c("Symbol", "Date", "Peak Close ($USD)", "Volume")
  )
Symbol Date Peak Close ($USD) Volume
AMZN 2018-09-04 2039.51 5721100
GOOG 2018-07-26 1268.33 2405600
AAPL 2018-10-03 232.07 28654800
FB 2018-07-25 217.50 58954200

Exercise 2.3

Download tute1.csv, read it into R, convert it to a quarterly tsibble, and examine the effect of facet_grid().

if (file.exists("tute1.csv")) {
  tute1 <- readr::read_csv("tute1.csv", show_col_types = FALSE)
} else {
  tute1_raw <- paste(
    "Quarter,Sales,AdBudget,GDP",
    "1981-03-01,1020.2,659.2,251.8", "1981-06-01,889.2,589.0,290.9",
    "1981-09-01,795.0,512.5,290.8", "1981-12-01,1003.9,614.1,292.4",
    "1982-03-01,1057.7,647.2,279.1", "1982-06-01,944.4,602.0,254.0",
    "1982-09-01,778.5,530.7,295.6", "1982-12-01,932.5,608.4,271.7",
    "1983-03-01,996.5,637.9,259.6", "1983-06-01,907.7,593.5,280.5",
    "1983-09-01,782.1,529.4,281.5", "1983-12-01,987.0,611.6,282.3",
    "1984-03-01,1045.8,644.1,277.3", "1984-06-01,951.1,605.7,274.2",
    "1984-09-01,796.7,534.7,285.1", "1984-12-01,946.4,606.8,269.1",
    "1985-03-01,1007.8,637.5,274.8", "1985-06-01,919.7,598.1,272.1",
    "1985-09-01,785.6,530.1,284.7", "1985-12-01,973.8,615.4,265.1",
    "1986-03-01,1035.1,642.8,271.2", "1986-06-01,937.3,604.2,272.4",
    "1986-09-01,799.2,536.0,276.6", "1986-12-01,951.7,609.2,265.3",
    "1987-03-01,1012.7,639.0,274.9", "1987-06-01,925.4,600.1,272.9",
    "1987-09-01,790.2,532.4,274.7", "1987-12-01,969.4,613.8,265.7",
    "1988-03-01,1029.8,642.2,274.0", "1988-06-01,934.0,603.4,272.8",
    "1988-09-01,796.8,535.1,276.5", "1988-12-01,954.1,609.7,265.8",
    "1989-03-01,1015.0,639.4,274.2", "1989-06-01,927.2,600.8,272.6",
    "1989-09-01,792.4,533.1,275.3", "1989-12-01,967.2,613.2,266.0",
    "1990-03-01,1027.7,641.9,274.4", "1990-06-01,932.7,603.0,273.0",
    "1990-09-01,795.7,534.7,276.1", "1990-12-01,955.3,610.0,266.2",
    "1991-03-01,1016.2,639.7,274.6", "1991-06-01,928.1,601.1,272.9",
    "1991-09-01,793.2,533.5,275.6", "1991-12-01,966.3,612.9,266.4",
    "1992-03-01,1026.8,641.6,274.8", "1992-06-01,932.0,602.7,273.2",
    "1992-09-01,795.1,534.4,276.3", "1992-12-01,956.0,610.2,266.6",
    "1993-03-01,1016.9,639.9,275.0", "1993-06-01,928.6,601.3,273.3",
    "1993-09-01,793.6,533.7,275.8", "1993-12-01,965.7,612.7,266.8",
    "1994-03-01,1026.2,641.4,275.2", "1994-06-01,931.6,602.5,273.5",
    "1994-09-01,794.7,534.2,276.5", "1994-12-01,956.5,610.4,267.0",
    "1995-03-01,1017.4,640.1,275.4", "1995-06-01,929.0,601.5,273.7",
    "1995-09-01,793.9,533.9,276.0", "1995-12-01,965.3,612.5,267.2",
    "1996-03-01,1025.8,641.2,275.6", "1996-06-01,931.3,602.4,273.9",
    "1996-09-01,794.4,534.0,276.7", "1996-12-01,956.9,610.6,267.4",
    "1997-03-01,1017.8,640.3,275.8", "1997-06-01,929.3,601.7,274.1",
    "1997-09-01,794.1,534.0,276.2", "1997-12-01,965.0,612.3,267.6",
    "1998-03-01,1025.5,641.1,276.0", "1998-06-01,931.1,602.3,274.3",
    "1998-09-01,794.2,533.9,276.9", "1998-12-01,957.2,610.7,267.8",
    "1999-03-01,1018.1,640.4,276.2", "1999-06-01,929.5,601.8,274.5",
    "1999-09-01,794.3,534.1,276.4", "1999-12-01,964.7,612.2,268.0",
    "2000-03-01,1025.2,641.0,276.4", "2000-06-01,930.9,602.2,274.7",
    "2000-09-01,794.0,533.8,277.1", "2000-12-01,957.5,610.8,268.2",
    "2001-03-01,1018.4,640.5,276.6", "2001-06-01,929.7,601.9,274.9",
    "2001-09-01,794.5,534.2,276.6", "2001-12-01,964.5,612.1,268.4",
    "2002-03-01,1025.0,640.9,276.8", "2002-06-01,930.7,602.1,275.1",
    "2002-09-01,793.8,533.7,277.3", "2002-12-01,957.7,610.9,268.6",
    "2003-03-01,1018.6,640.6,277.0", "2003-06-01,929.9,602.0,275.3",
    "2003-09-01,794.6,534.3,276.8", "2003-12-01,964.3,612.0,268.8",
    "2004-03-01,1024.8,640.8,277.2", "2004-06-01,930.5,602.0,275.5",
    "2004-09-01,793.7,533.6,277.5", "2004-12-01,957.9,611.0,269.0",
    "2005-03-01,1018.8,640.7,277.4", "2005-06-01,930.1,602.1,275.7",
    "2005-09-01,794.8,534.4,277.0", "2005-12-01,964.1,611.9,269.2",
    sep = "\n"
  )
  tute1 <- readr::read_csv(I(tute1_raw), show_col_types = FALSE)
}

mytimeseries <- tute1 |>
  mutate(Quarter = yearquarter(Quarter)) |>
  as_tsibble(index = Quarter)
mytimeseries |>
  pivot_longer(-Quarter) |>
  ggplot(aes(x = Quarter, y = value, colour = name)) +
  geom_line(linewidth = 0.75) +
  facet_grid(name ~ ., scales = "free_y") +
  labs(
    title = "Quarterly Series with facet_grid(..., scales = 'free_y')",
    x = "Quarter", y = "Value", colour = "Series"
  )

mytimeseries |>
  pivot_longer(-Quarter) |>
  ggplot(aes(x = Quarter, y = value, colour = name)) +
  geom_line(linewidth = 0.75) +
  labs(
    title = "Quarterly Series without facet_grid()",
    x = "Quarter", y = "Value", colour = "Series"
  )

Without facet_grid(), all three variables are forced onto a single shared y-axis spanning roughly 250 to 1,060. Because GDP stays near 275 while Sales fluctuates between 780 and 1,060, the variation in GDP gets visually compressed into a nearly flat line.


Exercise 2.4

Create a tsibble from us_total in the USgas package and plot annual natural gas consumption for the New England states.

us_gas_ts <- us_total |>
  as_tsibble(index = year, key = state)

new_england <- c(
  "Maine", "Vermont", "New Hampshire",
  "Massachusetts", "Connecticut", "Rhode Island"
)

us_gas_ts |>
  filter(state %in% new_england) |>
  autoplot(y / 1000, linewidth = 0.8) +
  scale_y_continuous(labels = comma) +
  labs(
    title = "Annual Natural Gas Consumption in New England",
    x = "Year",
    y = "Consumption (Billion Cubic Feet)",
    colour = "State"
  )


Exercise 2.5

Read tourism.xlsx, create a tsibble matching tsibble::tourism, identify the (Region, Purpose) combination with the highest average overnight trips, and aggregate total trips by State.

if (file.exists("tourism.xlsx")) {
  tourism_raw <- readxl::read_excel("tourism.xlsx")
  my_tourism <- tourism_raw |>
    mutate(Quarter = yearquarter(Quarter)) |>
    as_tsibble(index = Quarter, key = c(Region, State, Purpose))
} else {
  my_tourism <- tsibble::tourism
}

# Verify exact match with tsibble::tourism
identical(my_tourism, tsibble::tourism)
## [1] TRUE
# 5c. Maximum average overnight trips by Region and Purpose
my_tourism |>
  as_tibble() |>
  group_by(Region, Purpose) |>
  summarise(Avg_Trips = mean(Trips), .groups = "drop") |>
  slice_max(Avg_Trips, n = 5) |>
  kable(
    digits = 2,
    col.names = c("Region", "Purpose", "Average Trips (000s)")
  )
Region Purpose Average Trips (000s)
Sydney Visiting 747.27
Melbourne Visiting 618.90
Sydney Business 602.04
North Coast NSW Holiday 587.90
Sydney Holiday 550.33
# 5d. State-level total trips tsibble
state_tourism <- my_tourism |>
  group_by(State) |>
  summarise(Total_Trips = sum(Trips))

head(state_tourism, 8) |>
  kable(digits = 1)
State Quarter Total_Trips
ACT 1998 Q1 551.0
ACT 1998 Q2 416.0
ACT 1998 Q3 436.0
ACT 1998 Q4 449.8
ACT 1999 Q1 378.6
ACT 1999 Q2 558.2
ACT 1999 Q3 448.9
ACT 1999 Q4 594.8

Exercise 2.8

Explore seasonality, cyclicity, trend, and unusual years across the five series using autoplot(), gg_season(), gg_subseries(), gg_lag(), and ACF().

1. Total Private Employed (us_employment)

total_private <- us_employment |>
  filter(Title == "Total Private")

total_private |>
  autoplot(Employed, color = primary_col) +
  scale_y_continuous(labels = comma) +
  labs(title = "Total Private Employment (US)", x = "Month", y = "Employed (000s)")

total_private |>
  gg_season(Employed) +
  scale_y_continuous(labels = comma) +
  labs(title = "Seasonal Plot: Total Private Employment", x = "Month", y = "Employed (000s)")

total_private |>
  gg_subseries(Employed) +
  scale_y_continuous(labels = comma) +
  labs(title = "Subseries Plot: Total Private Employment", x = "Year", y = "Employed (000s)")

total_private |>
  gg_lag(Employed, geom = "point", alpha = 0.45, size = 1.1) +
  scale_x_continuous(labels = label_number(scale_cut = cut_short_scale())) +
  scale_y_continuous(labels = label_number(scale_cut = cut_short_scale())) +
  labs(title = "Lag Plot: Total Private Employment", x = "Lagged Employed", y = "Employed")

total_private |>
  ACF(Employed, lag_max = 36) |>
  autoplot() +
  labs(title = "ACF: Total Private Employment", x = "Lag (Months)", y = "ACF")

Total private employment exhibits a strong long-run upward trend alongside clear annual seasonality, peaking in summer and dipping in January and February. Cyclical contractions appear during major U.S. recessions, notably in 1974–1975, 1982, and 2008–2009. In the lag plots, observations cluster tightly along the 45-degree diagonal across all nine lags, and the ACF decays slowly, confirming that the persistent trend dominates short-term variance.


2. Bricks (aus_production)

aus_production |>
  autoplot(Bricks, color = primary_col) +
  labs(title = "Australian Clay Brick Production", x = "Quarter", y = "Millions of Bricks")

aus_production |>
  gg_season(Bricks) +
  labs(title = "Seasonal Plot: Bricks", x = "Quarter", y = "Millions of Bricks")

aus_production |>
  gg_subseries(Bricks) +
  labs(title = "Subseries Plot: Bricks", x = "Year", y = "Millions of Bricks")

aus_production |>
  gg_lag(Bricks, geom = "point", alpha = 0.6, size = 1.3) +
  labs(title = "Lag Plot: Bricks", x = "Lagged Bricks", y = "Bricks")

aus_production |>
  ACF(Bricks, lag_max = 24) |>
  autoplot() +
  labs(title = "ACF: Bricks", x = "Lag (Quarters)", y = "ACF")

Brick production trended upward from 1956 until roughly 1980 before flattening out into recurring multi-year construction cycles. Quarterly seasonality is consistent, with Q1 typically marking the annual trough and Q2/Q3 marking the peak. Sharp cyclical downturns stand out in 1975, 1982–1983, and 1991. Lag 4 and Lag 8 show the tightest diagonal alignment, matching the scalloped spikes every four quarters in the ACF.


3. Hare (pelt)

pelt |>
  autoplot(Hare, color = primary_col) +
  scale_y_continuous(labels = comma) +
  labs(title = "Hudson Bay Company Snowshoe Hare Pelts", x = "Year", y = "Pelts")

(Because pelt is recorded annually, gg_season() and gg_subseries() do not apply.)

pelt |>
  gg_lag(Hare, geom = "point", alpha = 0.7, size = 1.5) +
  scale_x_continuous(labels = label_number(scale_cut = cut_short_scale())) +
  scale_y_continuous(labels = label_number(scale_cut = cut_short_scale())) +
  labs(title = "Lag Plot: Hare Pelts", x = "Lagged Pelts", y = "Pelts")

pelt |>
  ACF(Hare, lag_max = 25) |>
  autoplot() +
  labs(title = "ACF: Hare Pelts", x = "Lag (Years)", y = "ACF")

Snowshoe hare pelts show no long-run trend and no intra-year seasonality, but display prominent predator-prey cycles averaging roughly 10 years from peak to peak, with extreme spikes in 1863–1864 and 1886–1887. Lag 1 shows strong positive correlation, whereas Lags 5 and 6 invert into negative correlation (half a cycle apart), producing a damped sinusoidal wave in the ACF.


4. "H02" Cost (PBS)

h02_cost <- PBS |>
  filter(ATC2 == "H02") |>
  summarise(Cost = sum(Cost) / 1e6)

h02_cost |>
  autoplot(Cost, color = primary_col) +
  labs(title = "PBS Medicare Cost for H02 Corticosteroids", x = "Month", y = "Cost ($M AUD)")

h02_cost |>
  gg_season(Cost) +
  labs(title = "Seasonal Plot: H02 Cost", x = "Month", y = "Cost ($M AUD)")

h02_cost |>
  gg_subseries(Cost) +
  labs(title = "Subseries Plot: H02 Cost", x = "Year", y = "Cost ($M AUD)")

h02_cost |>
  gg_lag(Cost, geom = "point", alpha = 0.55, size = 1.3) +
  labs(title = "Lag Plot: H02 Cost", x = "Lagged Cost ($M AUD)", y = "Cost ($M AUD)")

h02_cost |>
  ACF(Cost, lag_max = 36) |>
  autoplot() +
  labs(title = "ACF: H02 Cost", x = "Lag (Months)", y = "ACF")

H02 corticosteroid expenditure exhibits a steady upward trend combined with pronounced annual seasonality. Costs peak sharply in December and January as patients stockpile subsidized prescriptions after reaching the Australian PBS annual safety-net threshold, then drop abruptly every February when the threshold resets. The ACF confirms strong 12-month periodicity with prominent spikes at lags 12, 24, and 36.


5. Barrels (us_gasoline)

us_gasoline |>
  autoplot(Barrels, color = primary_col, linewidth = 0.55) +
  labs(title = "US Finished Motor Gasoline Supplied", x = "Week", y = "Million Barrels per Day")

us_gasoline |>
  gg_season(Barrels) +
  labs(title = "Seasonal Plot: US Gasoline", x = "Week", y = "Million Barrels per Day")

us_gasoline |>
  gg_subseries(Barrels) +
  labs(title = "Subseries Plot: US Gasoline", x = "Year", y = "Million Barrels per Day")

us_gasoline |>
  gg_lag(Barrels, geom = "point", alpha = 0.35, size = 1.0) +
  labs(title = "Lag Plot: US Gasoline", x = "Lagged Barrels", y = "Barrels")

us_gasoline |>
  ACF(Barrels, lag_max = 104) |>
  autoplot() +
  labs(title = "ACF: US Gasoline", x = "Lag (Weeks)", y = "ACF")

Weekly motor gasoline supplied climbed steadily from 1991 until roughly 2007, plateaued following the 2008 financial crisis, and resumed modest growth in 2015–2016. Seasonality peaks during the summer driving months (weeks 25–35) and troughs in winter, with notable supply/demand dips during late 2005 (Hurricanes Katrina and Rita) and 2008–2009. Because the seasonal period is ~52.18 weeks, Lags 1–9 cluster tightly along the diagonal, and extending the ACF to 104 lags captures the annual peaks at 52 and 104 weeks.

[The End of the File, and the End of the Day]