1  Introduction

1.1 What is Time Series Data?

Time series data consists of observations recorded over time, usually at regular intervals (e.g., daily, monthly, yearly). Some examples include:

  • Monthly rainfall totals
  • Daily COVID-19 cases
  • Hourly temperature readings
  • Yearly population counts

What makes time series special is that time is not just a variable, It carries important structure and dependencies. What happened today can depend on what happened yesterday, last month or even last year. This temporal uniqueness is what makes time series very powerful.

1.2 Why Time Series Analysis Matters

Time series analysis helps us to understand the past, monitor the present, and predict the future. Some real world examples are in:

  • Public Health: Forecasting disease outbreaks or hospital admissions.
  • Finance: Predicting stock prices, currency exchange rates, or sales revenue.
  • Environmental Science: Analysing temperature trends or rainfall patterns for climate studies.
  • Engineering: Monitoring sensor data to detect faults or changes in performance.

In many cases, this analysis supports decision making — whether it is planning resources, anticipating risks, or detecting unusual patterns.

1.3 Traditional vs. Tidy Approach

Historically, time series in R has been handled using ts objects and packages like forecast. These are still useful, but they do not fit perfectly with the modern tidy philosophy; data frames, pipelines and consistent syntax.

We will follow the tidy time series workflow using the tidyverse ecosystem. The workflow uses packages from the tidyverts ecosystem:

  • tsibble: A tidy data structure for time series (like a tibble but with special time handling).
  • fable: For forecasting models (ARIMA, ETS, etc) in a tidy way.
  • feasts: For exploratory analysis (seasonal plots, decomposition, autocorrelation).
  • lubridate: For working with dates and times.
  • ggplot2: For beautiful and flexible visualisations.
  • dplyr/tidyr/tibble: For general data wrangling.

1.3.1 The Tidy Time Series Workflow

The tidyverse ecosystem in R emphasises clean, readable code and consistent data structure. For time series, the modern approach uses the tidyverts suite of packages. The typical flow we will follow includes:

  1. 🔧 Data Preparation: Load and tidy data. Convert the data to a tsibble object so R knows how to handle time.
  2. 🔍 Data Exploration: Visualise trends, seasonality, patterns
  3. 🧩 Decomposing Time Series: Break down components (trends, seasonal, noise)
  4. 🖥️ Mode Fitting: Fit forecasting models (simple to advanced)
  5. 📈 Forecasting: Generate future predictions
  6. 📏 Model Evaluation: Check how good the forecasts are (accuracy assessment)

This workflow is clean, consistent and integrates smoothly with other tidyverse tools you might already know.

Tip

tidyverse: collection of packages designed for general data science

tidyverts: collection of packages specifically for time series analysis

They all follow the tidy philosophy, structure and grammar

1.3.2 Data We Will Use

We will start with some built-in datasets from the tsibbledata package so that you can follow along without downloading any external files yet.

Later, we shall also show how to use real-life datasets — for example, population growth or GDP growth data from Ghana, monthly sales data and other local datasets to make the examples relatable and practical. You can download all the custom datasets used here.

1.3.3 What You Will Need to Follow Along

  1. Basic R Knowledge: You should know how to load packages, run functions, and work with data frames.
  2. RStudio installed for a smooth workflow.
  3. Internet connection (for package installation and possible data download).

1.3.4 By the end of this book, you will be able to:

  • Handle date/time data with ease.
  • Explore time series visually and statistically.
  • Build forecasts using tidyverse-style functions.
  • Apply your skills to your own datasets in research or work.