2021/6/6 https://lms.monash.edu/pluginfile.php/12413905/mod_resource/content/0/3-decomposition.R
library(fpp3)
## GDP ————————————————————————–
global_economy %>%
filter(Country == “Australia”) %>%
autoplot(GDP)
global_economy %>%
filter(Country == “Australia”) %>%
autoplot(GDP / Population)
## Print retail adjusted by CPI ————————————————–
print_retail <- aus_retail %>%
filter(Industry == “Newspaper and book retailing”) %>%
group_by(Industry) %>% # Just to keep the key in there
index_by(Year = year(Month)) %>%
summarise(Turnover = sum(Turnover))
aus_CPI <- global_economy %>%
filter(Code == “AUS”) %>%
select(CPI)
print_retail_adj <- print_retail %>%
left_join(aus_CPI, by = “Year”) %>%
mutate(Adj_turnover = Turnover / CPI*100) %>%
pivot_longer(c(Turnover, Adj_turnover),
names_to = “Type”, values_to = “Turnover”
) %>%
select(Turnover,-Industry)
# Plot both on same graph
print_retail_adj %>%
ggplot(aes(x = Year, y = Turnover, col=Type)) +
geom_line() +
labs(title = “Turnover: Australian print media industry”,
y = “$AU”)
# Use faceting
print_retail_adj %>%
mutate(name = factor(Type,
levels=c(“Turnover”,”Adj_turnover”))) %>%
ggplot(aes(x = Year, y = Turnover)) +
geom_line() +
facet_grid(name ~ ., scales = “free_y”) +
labs(title = “Turnover: Australian print media industry”,
y = “$AU”)
# Return to slides to comment
## Australian food retail ——————————————————–
food <- aus_retail %>%
filter(Industry == “Food retailing”) %>%
summarise(Turnover = sum(Turnover)) #summed over states
food %>% autoplot(Turnover) +
labs(y = “Turnover ($AUD)”)
food %>% autoplot(sqrt(Turnover)) +
labs(y = “Square root turnover”)
food %>% autoplot(log(Turnover)) +
labs(y = “Log turnover”)
# Tell i
food %>%
https://lms.monash.edu/pluginfile.php/12413905/mod_resource/content/0/3-decomposition.R 1/3
2021/6/6 https://lms.monash.edu/pluginfile.php/12413905/mod_resource/content/0/3-decomposition.R
features(Turnover, features = guerrero)
food %>% autoplot(box_cox(Turnover, 0.0524)) +
labs(y = “Box-Cox transformed turnover”)
## US retail employment ———————————————————-
us_retail_employment <- us_employment %>%
filter(year(Month) >= 1990, Title == “Retail Trade”) %>%
select(-Series_ID)
us_retail_employment %>%
autoplot(Employed) +
xlab(“Year”) + ylab(“Persons (thousands)”) +
ggtitle(“Total employment in US retail”)
dcmp <- us_retail_employment %>%
model(stl = STL(Employed))
components(dcmp)
us_retail_employment %>%
autoplot(Employed, color=’gray’) +
autolayer(components(dcmp), trend, color=’red’) + #pull out the trend
xlab(“Year”) + ylab(“Persons (thousands)”) +
ggtitle(“Total employment in US retail”)
components(dcmp) %>% autoplot() + xlab(“Year”)
components(dcmp) %>% gg_subseries(season_year)
us_retail_employment %>%
autoplot(Employed, color=’gray’) +
autolayer(components(dcmp), season_adjust, color=’blue’) +
xlab(“Year”) + ylab(“Persons (thousands)”) +
ggtitle(“Total employment in US retail”)
#### MA ———–
us_retail_employment_ma <- us_retail_employment %>%
mutate(
`12-MA` = slider::slide_dbl(Employed, mean,
.before = 5, .after = 6, .complete = TRUE),
`2×12-MA` = slider::slide_dbl(`12-MA`, mean,
.before = 1, .after = 0, .complete = TRUE)
)
autoplot(us_retail_employment_ma, Employed, color = “gray”) +
autolayer(us_retail_employment_ma, vars(`2×12-MA`),
color = “#D55E00”) +
labs(y = “Persons (thousands)”,
title = “Total employment in US retail”)
#### Classical decomposition
us_retail_employment %>%
model(classical_decomposition(Employed, type = “additive”)) %>%
components() %>%
autoplot() + xlab(“Year”) +
ggtitle(“Classical additive decomposition of total US retail employment”)
##### STL decomposition
# Let’s start with the basic model
us_retail_employment %>%
https://lms.monash.edu/pluginfile.php/12413905/mod_resource/content/0/3-decomposition.R 2/3
2021/6/6 https://lms.monash.edu/pluginfile.php/12413905/mod_resource/content/0/3-decomposition.R
model(STL(Employed)) %>%
components() %>%
autoplot() +
ggtitle(“STL decomposition: US retail employment”)
# We saw this before
# Let’s play around with some parameters and let’s see what happens
# Let’s control the seasonal component
# season(window=13) – Default setting – no science behind
# set season(window=91) equivalent to “periodic”
# set season(window=13)
# set season(window=9)
# set season(window=3)
us_retail_employment %>%
model(STL(Employed ~ season(window=”periodic”))) %>%
components() %>%
autoplot() +
ggtitle(“STL decomposition: US retail employment”)
# Let’s control the trend
# Default is complex automated algorithm depends on size of
# season and noise – works fairly well so we usually leave it alone
# set trend(window=5) – notice the remainder
# set trend(window=99, 999)
# Default setting for monthly data is 21
us_retail_employment %>%
model(STL(Employed ~ season(window=9))) %>%
components() %>%
autoplot() +
ggtitle(“STL decomposition: US retail employment”)
# Robust option – make this true and any outliers should
# show up im remainder
us_retail_employment %>%
model(STL(Employed ~ season(window=9) + trend(window=21), robust=TRUE)) %>%
components() %>%
autoplot() +
ggtitle(“STL decomposition: US retail employment”)
#
https://lms.monash.edu/pluginfile.php/12413905/mod_resource/content/0/3-decomposition.R 3/3