Histogram Pretty

Histogram Pretty was built to identify reasonable histogram bucket sizing for charting applications. It tries to follow a combination of the R base graphics histogram logic and D3. I've chosen the Freedman and Diaconis (1981) Rule, which limits the influence of outliers in bucket width selection. This seems to significantly improve binning for dynamic charting where the data is often filtered into irregular subsets.

Usage

var vector = ... // vector extraction logic
var hist = histogram(vector);

Parameters

The vector should be an array of numeric values. Optionally, you may also pass an options object with one of two values:

• copy If truthy, will copy the array with vector.slice() prior to internally sorting the vector. It is true by default, but if mutability is acceptable can be set to true.

• pretty If truthy, will round bin sizing to a factor of 2, 5 or 10 within the relevant scale of the bin size. It is true by default. Otherwise the raw bin size as returned by the Freedman-Diaconis Rule will be returned.

Returns

The histogram function returns an object with three values:

• size - the bin width
• fun - a binning function. This is simply: function(d) { return h * Math.floor(d / h); } Where h is the size.
• tickRange will return a function to compute ticks for visual display on a chart axis. Some charting libraries (such as d3) do not take into account the discretization when identify axis ticks. This function will always break on a bin boundary. Call tickRange(n) with a requested number of ticks as appropriate for the chart display. The function will return approximately n ticks as an array of breaks in the units of the vector.

Example

See: http://jrideout.github.io/histogram-pretty