Classification methods

Numeric, H3, and raster (numeric and tinted hillshade) visualizations all turn a range of values into discrete classes (or a smooth gradient) using the steps field in the config block. The classification method decides where the breaks between classes fall.

Methods

Method	What it does	Best for
continuous	Smooth gradient, no discrete classes	Proportional symbols, smooth color ramps
jenks	Natural breaks — minimizes variance within each class	Most cases; this is the default
quantiles	Equal number of features in each class	Balanced representation, evenly distributed data
equal-intervals	Equal-sized value bins	When the value ranges themselves matter
stddev	Standard deviations from the mean	Data with a near-normal distribution
geo-intervals	Geometric / geographic intervals	Spatial distribution, multiplicative data

Specifying steps

There are two ways to define steps.

Shortcut (recommended when you don't know the data's distribution). Name a method and a class count and let Felt compute the break points from the data:

{
  "config": {
    "numericAttribute": "population_density",
    "steps": {"type": "jenks", "count": 5}
  }
}

For a smooth gradient, use the continuous shortcut:

{
  "config": {
    "numericAttribute": "temperature",
    "steps": {"type": "continuous"}
  }
}

Explicit break points (when you know the values you want). Provide the break points directly:

{
  "config": {
    "numericAttribute": "magnitude",
    "steps": [0, 3, 5, 7, 9]
  }
}

How breaks map to colors and sizes

• Classed: N break points produce N − 1 classes. An explicit color or size array must therefore have exactly N − 1 entries. (A palette shortcut such as @galaxy is expanded for you.)

• Continuous: provide 2 or more colors and Felt interpolates between them across the full value range. For proportional symbols, give size: [min, max] and sizes scale smoothly between those two values.

For choropleths and other classed thematic maps, 5–7 classes usually reads best. For right-skewed data (common with population or income), prefer quantiles or jenks over equal-intervals. When you have the data's min/max on hand, use them to inform the method and breaks.

See Numeric visualizations for full color and size examples, and Colors & palettes for palette choices.