How to choose a good chart

2013-04-11 21:58:40

Visualization is a very import part of data science. Andrew Abela posted a way for how to choose a good chart to help visualize data.

Here are some summaries:

For comparisons:

use Bar Chart, horizontal for small variables, vertical for more variables.

For trending information:

use Line Chart, but do no use for categorical data.
bank to 45 degree.

For scatterplots:

avoid over-plotting.
avoid 3D scatter plots.

For compositions:

Pie Chart, Donut Chart or Stacked Bar Chart.

For distributions:

Histograms or Density Plots for single variable.
Heat map for 2D data.