Introduction

For almost a decade, the forecast package has been a rock-solid framework for time series forecasting. However, within the last year or so an official updated version has been released named fable which now follows tidy methods as opposed to base R.

More recently, modeltime has been released and this also follows tidy methods. However, it is strictly used for modeling. For data manipulation and visualization, the timetk package will be used which is written by the same author as modeltime.

The following is a code comparison of various time series visualizations between these frameworks: fpp2, fpp3 and timetk.

#Load Libraries

# Load libraries
library(fpp2)         # An older forecasting framework
library(fpp3)         # A newer tidy forecasting framework
library(timetk)       # An even newer tidy forecasting framework
library(tidyverse)    # Collection of data manipulation tools
library(tidyquant)    # Business Science ggplot theme
library(cowplot)      # A ggplot add-on for arranging plots

TS vs tsibble

the base ts object is used by forecast & fpp2
The special tsibble object is used by fable & fpp3
The standard tibble object is used by timetk & modeltime

1. Load Time Series Data

we will utilize a dataset containing quarterly production values of certain commodities in Australia.

# Quarterly Australian production data as tibble
aus <- tsibbledata::aus_production %>% as_tibble()
# Check structure
aus %>% str()

## tibble [218 × 7] (S3: tbl_df/tbl/data.frame)
##  $ Quarter    : qtr [1:218] 1956 Q1, 1956 Q2, 1956 Q3, 1956 Q4, 1957 Q1, 1957 Q2, 1957...
##  $ Beer       : num [1:218] 284 213 227 308 262 228 236 320 272 233 ...
##  $ Tobacco    : num [1:218] 5225 5178 5297 5681 5577 ...
##  $ Bricks     : num [1:218] 189 204 208 197 187 214 227 222 199 229 ...
##  $ Cement     : num [1:218] 465 532 561 570 529 604 603 582 554 620 ...
##  $ Electricity: num [1:218] 3923 4436 4806 4418 4339 ...
##  $ Gas        : num [1:218] 5 6 7 6 5 7 7 6 5 7 ...

2 fpp2 Method: From tibble to ts

# Convert tibble to time series object
aus_prod_ts <- ts(aus[, 2:7],  # Choose columns
                    start = c(1956, 1),  # Choose start date
                    end = c(2010, 2),    # Choose end date
                    frequency = 4)       # Choose frequency per yr
# Check it out
aus_prod_ts %>% tail()

##         Beer Tobacco Bricks Cement Electricity Gas
## 2009 Q1  415      NA     NA   1963       58368 196
## 2009 Q2  398      NA     NA   2160       57471 238
## 2009 Q3  419      NA     NA   2325       58394 252
## 2009 Q4  488      NA     NA   2273       57336 210
## 2010 Q1  414      NA     NA   1904       58309 205
## 2010 Q2  374      NA     NA   2401       58041 236

#3 fpp3 Method: From ts to tsibble

3.1 Pivot Wide

# Convert ts to tsibble and keep wide format
aus_prod_tbl_wide <- aus_prod_ts %>%    # TS object
  as_tsibble(index = "index",           # Set index column
             pivot_longer = FALSE)      # Wide format

## 3.2 Pivot Long Convert ts to tsibble and pivot to long format

aus_prod_tbl_long <- aus_prod_ts %>%    # TS object
  as_tsibble(index = "index", pivot_longer = TRUE) # Long format

#4 timetk Method: From tsibble/ts to tibble ## 4.1 Pivot Wide

# Convert tsibble to tibble, keep wide format
aus <- tsibbledata::aus_production %>% 
    tk_tbl() %>%
    mutate(Quarter = as_date(as.POSIXct.Date(Quarter)))

4.2 Pivot Long

# Quarterly Australian production data to long format
aus_long <- aus %>% 
    rename(date = Quarter) %>%
    pivot_longer(
        cols = c("Beer","Tobacco","Bricks",
                 "Cement","Electricity","Gas"))

5. Time Series Plots

When analyzing time series plots, look for the following patterns:

-Trend: A long-term increase or decrease in the data; a “changing direction”.

_Seasonality: A seasonal pattern of a fixed and known period. If the frequency is unchanging and associated with some aspect of the calendar, then the pattern is seasonal.

-Cycle: A rise and fall pattern not of a fixed frequency. If the fluctuations are not of a fixed frequency then they are cyclic.

-Seasonal vs Cyclic: Cyclic patterns are longer and more variable than seasonal patterns in general.

5.1 fpp2 Method: Plot Multiple Series On Same Axes

# Using fpp2
aus_prod_ts %>%               # TS object
  autoplot(facets=FALSE)      # No facetting

quaterly production of selected quantities using fpp2

5.2 fpp3 Method: Plot Multiple Series On Same Axes

# Using fpp3
aus_prod_tbl_long %>%    # Data in long format
  autoplot(value)

quaterly production of selected quantities using fpp3

5.3 ggplot Method: Plot Multiple Series On Same Axes

plotting multiple plots on the same axes has not been implemented into timetk. hence, we Use ggplot.

# Using ggplot
aus_long %>%
    ggplot(aes(date, value, group = name, color = name)) +
    geom_line()

quaterly production of selected quantities using ggplot

5.4 fpp2 Method: Plot Multiple Series On Separate Axes

# Using fpp2
aus_prod_ts %>%  
  autoplot(facets=TRUE)   # With facetting

quaterly production of selected quantities using fpp2

5.5 fpp3 Method: Plot Multiple Series On Separate Axes

# Using fpp3
aus_prod_tbl_long %>%
  ggplot(aes(x = index, y = value, group = key)) + 
  geom_line() + 
  facet_grid(vars(key), scales = "free_y")   # With facetting

quaterly production of selected quantities using fpp3

5.6 timetk Method: Plot Multiple Series On Separate Axes

# Using timetk
aus_long %>% 
    plot_time_series(
        .date_var = date,
        .value = value,
        .facet_vars = c(name), # Group by these columns
        .color_var = name, 
        .interactive = FALSE,
        .legend_show = FALSE
    )

quaterly production of selected quantities using timetk

Seasonal Plots

Use seasonal plots for identifying time periods in which the patterns change.

## 6.1 fpp2 Method: Plot Individual Seasons

# Monthly plot of anti-diabetic scripts in Australia 
a1 <- a10 %>%
  autoplot() 
# Seasonal plot
a2 <- a10 %>% 
  ggseasonplot(year.labels.left = TRUE,   # Add labels
               year.labels = TRUE) 
# Arrangement of plots
plot_grid(a1, a2, ncol=1, rel_heights = c(1, 1.5))

Monthly plot of anti-diabetic scripts

6.2 fpp3 Method: Plot Individual Seasons

# Monthly plot of anti-diabetic scripts in Australia
a1 <- a10 %>%
  as_tsibble() %>%
  autoplot(value)
# Seasonal plot
a2 <- a10 %>%
  as_tsibble() %>%
  gg_season(value, labels="both")   # Add labels
# Arrangement of plots
plot_grid(a1, a2, ncol=1, rel_heights = c(1, 1.5))

Monthly plot of anti-diabetic scripts

6.3 ggplot Method: Plot Individual Seasons

Note that seasonal plots have not been implemented into timetk .hence, Use ggplot to write:

# Convert ts to tibble
a10_tbl <- fpp2::a10 %>%
    tk_tbl()
# Monthly plot of anti-diabetic scripts in Australia 
a1 <- a10_tbl %>% 
    plot_time_series(
        .date_var = index,
        .value    = value,
        .smooth   = TRUE,
        .interactive = FALSE,
        .title = "Monthly anti-diabetic scripts in Australia"
    )
# New time-based features to group by
a10_tbl_add <- a10_tbl %>% 
    mutate( 
        month = factor(month(index, label = TRUE)),  # Plot this
        year = factor(year(index))  # Grouped on y-axis
    )
# Seasonal plot
a2 <- a10_tbl_add %>%
    ggplot(aes(x = month, y = value, 
               group = year, color = year)) + 
    geom_line() + 
    geom_text(
        data = a10_tbl_add %>% filter(month == min(month)),
        aes(label = year, x = month, y = value),
        nudge_x = -0.3) + 
    geom_text(
        data = a10_tbl_add %>% filter(month == max(month)),
        aes(label = year, x = month, y = value),
        nudge_x = 0.3) + 
    guides(color = FALSE)
# Arrangement of plots
plot_grid(a1, a2, ncol=1, rel_heights = c(1, 1.5))

Monthly plot of anti-diabetic scripts

7. Lag Plots

Use lag plots to check for randomness.

7.1 fpp2 Method: Plot Multiple Lags

# Plot of non-seasonal oil production in Saudi Arabia
o1 <- fpp2::oil %>%
  autoplot()
# Lag plot of non-seasonal oil production
o2 <- gglagplot(oil, do.lines = FALSE)
# Plot both
plot_grid(o1, o2, ncol=1, rel_heights = c(1,2))

Annual oil production in Saudi Arabia

7.2 fpp3 Method: Plot Multiple Lags

# Plot of non-seasonal oil production
o1 <- oil %>%
  as_tsibble() %>%
  autoplot(value)
# Lag plot of non-seasonal oil production
o2 <- oil %>%
  as_tsibble() %>%
  gg_lag(y=value, geom = "point") 
# Plot it
plot_grid(o1, o2, ncol=1, rel_heights = c(1,2))

Annual oil production in Saudi Arabia

7.3 timetk Method: Plot Multiple Lags

# Convert to tibble and create lag columns
oil_lag_long <- oil %>%
    tk_tbl(rename_index = "year") %>%
    tk_augment_lags(      # Add 9 lag columns of data
        .value = value, 
        .names = "auto", 
        .lags = 1:9) %>%
    pivot_longer(         # Pivot from wide to long
        names_to = "lag_id", 
        values_to = "lag_value", 
        cols = value_lag1:value_lag9) # Exclude year & value

then the plot:

# Time series plot
o1 <- oil %>%
    tk_tbl(rename_index = "year") %>%  
    mutate(year = ymd(year, truncated = 2L)) %>%  
    plot_time_series(
        .date_var = year, 
        .value = value,
        .interactive = FALSE)
# timetk Method: Plot Multiple Lags
o2 <- oil_lag_long %>%
    plot_time_series(
        .date_var = value,     # Use value instead of date
        .value = lag_value,    # Use lag value to plot against
        .facet_vars = lag_id,  # Facet by lag number
        .facet_ncol = 3,
        .interactive = FALSE, 
        .smooth = FALSE,
        .line_alpha = 0,      
        .legend_show = FALSE,
        .facet_scales = "fixed"
        ) + 
    geom_point(aes(colour = lag_id)) + 
    geom_abline(colour = "gray", linetype = "dashed") 
# Plot it
plot_grid(o1, o2, ncol=1, rel_heights = c(1,2))

Annual oil production in Saudi Arabia

8. Autocorrelation Function Plots

The autocorrelation function measures the linear relationship between lagged values of a time series. The partial autocorrelation function measures the linear relationship between the correlations of the residuals.

ACF

-Visualizes how much the most recent value of the series is correlated with past values of the series (lags) -If the data has a trend, then the autocorrelations for small lags tend to be positive and large because observations nearby in time are also nearby in size -If the data are seasonal, then the autocorrelations will be larger for seasonal lags at multiples of seasonal frequency than other lags

PACF

-Visualizes whether certain lags are good for modeling or not; useful for data with a seasonal pattern Removes dependence of lags on other lags by using the correlations of the residuals

8.1 fpp2 Method: Plot ACF + PACF

# ACF plot 
o1 <- ggAcf(oil, lag.max = 20)
# PACF plot
o2 <- ggPacf(oil, lag.max = 20)
# Plot both
plot_grid(o1, o2, ncol = 1)

Auto Correlation Function

8.2 fpp3 Method: Plot ACF + PACF

# Convert to tsibble
oil_tsbl <- oil %>% as_tsibble()
# ACF Plot
o1 <- oil_tsbl %>%
  ACF(lag_max = 20) %>%
  autoplot()
# PACF Plot
o2 <- oil_tsbl %>%
  PACF(lag_max = 20) %>%
  autoplot() 
# Plot both
plot_grid(o1, o2, ncol = 1)

Auto Correlation Function

7.3 timetk Method: Plot ACF & PACF

# Using timetk
oil %>%
    tk_tbl(rename_index = "year") %>%
    plot_acf_diagnostics(
        .date_var = year,
        .value    = value,
        .lags     = 20,
        .show_white_noise_bars = TRUE, 
        .interactive = FALSE
    )

Auto Correlation Function

Summary

As with all things in life, there are good and bad sides to using any of these three forecasting frameworks for visualizing time series. All three have similar functionality as it relates to visualizations.

fpp2

-Code requires minimal parameters -Uses basets format -Uses ggplot for visualizations -Mostly incompatible with tidyverse for data manipulation -No longer maintained except for bug fixes

fpp3

-Code requires minimal parameters -Uses proprietary tsibble format with special indexing tools -Uses ggplot for visualizations -Mostly compatible with tidyverse for data manipulation; tsibble may cause issues -Currently maintained

timetk

-Code requires multiple parameters but provides more granularity -Uses standard tibble format -Uses ggplot and plotly for visualizations -Fully compatible with tidyverse for data manipulation -Currently maintained

Badawi Aminu Muhammed

Foreasting Time Series Data in R with tidy

Badawi Amin Muhammed

2025-11-22

Introduction

TS vs tsibble

1. Load Time Series Data

2 fpp2 Method: From tibble to ts

3.1 Pivot Wide

4.2 Pivot Long

5. Time Series Plots

5.1 fpp2 Method: Plot Multiple Series On Same Axes

5.2 fpp3 Method: Plot Multiple Series On Same Axes

5.3 ggplot Method: Plot Multiple Series On Same Axes

5.4 fpp2 Method: Plot Multiple Series On Separate Axes

5.5 fpp3 Method: Plot Multiple Series On Separate Axes

5.6 timetk Method: Plot Multiple Series On Separate Axes

Seasonal Plots

6.2 fpp3 Method: Plot Individual Seasons

6.3 ggplot Method: Plot Individual Seasons

7. Lag Plots

7.1 fpp2 Method: Plot Multiple Lags

7.2 fpp3 Method: Plot Multiple Lags

7.3 timetk Method: Plot Multiple Lags

8. Autocorrelation Function Plots

8.1 fpp2 Method: Plot ACF + PACF

8.2 fpp3 Method: Plot ACF + PACF

7.3 timetk Method: Plot ACF & PACF

Summary