Statistical Fits & Clustering

Regression and LOESS fits with confidence bands, and hierarchical or k-means clustering — built into the engine.

Introduction

CanvasXpress computes several common statistics in the chart itself, so they travel with the figure and work the same from JavaScript, R, Python, notebooks and dashboards. This guide covers trend fits on scatter plots and clustering on heatmaps.

Regression & LOESS Fits

On a 2D scatter plot, set showRegressionFit for a parametric fit (choose the model with regressionType) or showLoessFit for a local regression. Add showConfidenceIntervals to shade the interval; fitLineColor and confidenceIntervalColor set the colors.

var cXFit = new CanvasXpress("canvasFit", fitData, {
  graphType: "Scatter2D",
  showRegressionFit: true,
  regressionType: "linear",
  showConfidenceIntervals: true,
  fitLineColor: "#0000AA"
});

Parameter	Description
`showRegressionFit`	Draw a parametric regression line
`regressionType`	Model for the regression fit (e.g. `linear`)
`showLoessFit`	Draw a local (LOESS) regression curve
`showConfidenceIntervals`	Shade the confidence band around the fit
`confidenceIntervalColor`	Color of the confidence band
`fitLineColor`	Color of the fitted line

Clustering Heatmaps

On a heatmap, set samplesClustered and/or variablesClustered to cluster the columns and rows; a dendrogram is drawn alongside the clustered axis. Control the computation with distance and linkage, or switch to k-means with kmeansSmpClusters / kmeansVarClusters (and maxIterations).

{
  graphType: "Heatmap",
  samplesClustered: true,
  variablesClustered: true,
  distance: "euclidean",
  linkage: "complete"
}

Clustering can also be toggled from the chart’s right-click menu and the Customizer, and the resulting order is saved with the configuration.

Scope

Trend fits apply to 2D scatter plots; forecasting a time series is a separate feature (see Forecasting).
Hierarchical and k-means clustering apply to heatmaps and the dendrograms derived from them.
All of these default to off, so existing charts are unchanged.