Introduction

Ever wondered how to compare different subsets of the same data visually? Enter Trellis plots (or small multiples)! They are a series of similar plots displaying different subsets of the same data, making comparisons a breeze.

In Vega-Lite, you can create Trellis plots through two main methods:

1. The facet operator: For more flexibility and the ability to compose with other operators.
2. The facet, column, or row encoding channels: A handy shortcut for creating faceted plots quickly.

Let's dig in!

Using the Facet Operator

The facet operator is like the Swiss Army knife of faceting in Vega-Lite. It offers great flexibility but requires a bit more setup.

Facet Specification

To create a faceted view, define how the data should be split in the facet property and specify the layout in the spec property.

{
  "facet": {
    // Define your facets here
  },
  "spec": {
    // Define your visualization specs here
  }
}

The facet property includes specific settings like bin, field, timeUnit, type, and header. These settings tell Vega-Lite how to divide your data into different plots.

Example: Row-Facet

Imagine you're working with a dataset of cars, and you want to create histograms for the horsepower of cars based on their origin. Here's how you can achieve that using the facet operator:

{
  "facet": {
    "row": { "field": "origin", "type": "nominal" }
  },
  "spec": {
    // Further specifications for the histogram
  }
}

This setup will create a vertical series of histograms for each car origin—Europe, Japan, and USA.

Example: Wrapped Facet

Want to see all your data facets in a grid? Use a wrapped facet!

{
  "facet": {
    "field": "origin",
    "type": "nominal",
    "columns": 3
  },
  "spec": {
    // Further specifications for the plots
  }
}

Using Facet, Row, and Column Encoding Channels

For a more straightforward approach, use Vega-Lite's encoding channels: facet, row, and column.

Row Facet (with Row Encoding)

This method is simpler and Vega-Lite will do the heavy lifting for you:

{
  "mark": "bar",
  "encoding": {
    "row": { "field": "origin", "type": "nominal" },
    "x": { "field": "horsepower", "type": "quantitative" }
  }
}

With this, Vega-Lite generates a similar row-wise faceted plot like the one we created using the facet operator.

Grid Facet

To produce a grid of small multiples using both row and column channels:

{
  "mark": "bar",
  "encoding": {
    "row": { "field": "origin", "type": "nominal" },
    "column": { "field": "cylinders", "type": "ordinal" },
    "x": { "field": "horsepower", "type": "quantitative" },
    "y": { "aggregate": "count", "type": "quantitative" }
  }
}

Wrapped Facet (with Facet Encoding)

A wrapped facet layout through encoding:

{
  "mark": "point",
  "encoding": {
    "facet": { "field": "origin", "type": "nominal", "columns": 3 },
    "x": { "field": "horsepower", "type": "quantitative" },
    "y": { "field": "weight", "type": "quantitative" }
  }
}

FAQ

What is a Trellis Plot and why should I use it?

A Trellis Plot, also known as a small multiple, is a series of similar plots shown together. They help in visually comparing different subsets of the same data, making patterns and differences easy to detect.

When should I use the facet operator instead of encoding channels?

Use the facet operator when you need more flexibility and control over your layout and want to combine it with other Vega-Lite operators. Encoding channels are excellent for quick, straightforward faceting.

Can I facet based on numerical fields?

You should not facet by unprocessed numerical fields. Instead, use binning for numerical fields or specific time units for temporal fields, ensuring proper partitioning of your data.

Conclusion

Creating faceted plots or Trellis plots in Vega-Lite is both powerful and flexible, catering to quick setups and detailed customizations. Whether you use the facet operator or encoding channels, you can easily visualize and compare different subsets of your data.

Happy Visualizing!