Skip to main content

Time Series

The Time Series module provides tools for exploring and analysing time series data. It is built on the modern tidyverts ecosystem, providing flexible handling of various time series formats and powerful visualisation capabilities.

tip

This is the recommended Time Series module. If you need access to the older version (pre-2023), see Time Series (Legacy).

Draft Documentation

This documentation was generated based on source code inspection and may not fully reflect the current software. Items to verify:

  • Exact menu paths and UI controls
  • Available plot types and their options
  • Supported time format patterns
  • Settings and their effects

Features​

  • Flexible time formats: Supports various time frequencies (yearly, quarterly, monthly, weekly, daily, hourly)
  • Multiple series: Handle panel data with "keys" to analyse multiple related time series
  • Modern decomposition: STL decomposition to separate trend, seasonal, and residual components
  • ARIMA forecasting: Automatic model selection for accurate predictions
  • Interactive plots: Built with ggplot2 for high-quality, customisable graphics

Getting Started​

Accessing the Module​

Go to Advanced → Time Series from the menu bar.

Preparing Your Data​

Your data should have:

  1. A time/date column - containing dates, timestamps, or time codes
  2. One or more measurement columns - the values to analyse over time
  3. (Optional) Key columns - categorical variables identifying different series in panel data

Time Information​

Select your time variable from the dropdown. iNZight will automatically detect the time format if possible. Supported formats include:

FormatExamples
Dates2023-01-15, 15/01/2023, Jan 2023
Quarters2023Q1, 2023-Q1, Q1 2023
Year-Month2023M01, 2023-01, Jan-2023
Year-Week2023W01, 2023-W01
Years2023, Y2023

Working with Multiple Series (Keys)​

If your data contains multiple time series (e.g., sales data for different regions), select the identifying variable(s) as Keys. This allows iNZight to:

  • Plot all series together for comparison
  • Decompose and forecast each series separately
  • Highlight specific series of interest

Plot Types​

Select the plot type from the Plot Type dropdown in the module panel.

Default Plot​

Shows the raw time series with an optional smoother overlay. Use this to:

  • Identify overall trends
  • Spot outliers or unusual patterns
  • Compare multiple series

Options:

  • Show smoother: Toggle trend line on/off
  • Smoothness: Adjust the smoothness of the trend (0-100)
  • Seasonally adjust series: Remove seasonal effects to see underlying trend

Decomposition Plot​

Breaks the time series into components:

  • Trend: The long-term direction of the data
  • Seasonal: Repeating patterns within each period
  • Remainder: What's left after removing trend and season

Choose between Additive (Season + Trend + Remainder = Data) or Multiplicative (Season × Trend × Remainder = Data) decomposition based on whether seasonal variation changes with the level of the series.

Seasonal Plot​

Displays the seasonal pattern by overlaying data from each period (year, week, etc.) on the same axes. This helps identify:

  • Consistent seasonal patterns
  • Changes in seasonal behaviour over time
  • Anomalous periods

Forecast Plot​

Generates predictions for future time points using ARIMA models. The plot shows:

  • Historical data
  • Fitted values
  • Future predictions with confidence intervals

Settings​

Seasonal Pattern​

  • Additive: Use when seasonal fluctuations are roughly constant regardless of the level
  • Multiplicative: Use when seasonal fluctuations grow proportionally with the level

Variable Types​

  • Numeric Variables: Standard time series of measurements
  • Categorical Variables: Track how category proportions change over time

Range Settings​

Adjust the time range to focus on specific periods of interest. This affects both the display and the data used for model fitting.

Tips​

  1. Start with the default plot to get an overview of your data
  2. Check decomposition to understand the underlying structure
  3. Use seasonal plots to verify seasonal patterns before forecasting
  4. Adjust smoothness - lower values show more detail, higher values show clearer trends
  5. For forecasting, ensure you have enough historical data (typically at least 2-3 complete seasonal cycles)

Additional Resources​