Skip to content

Latest commit

 

History

History
56 lines (40 loc) · 1.6 KB

README.md

File metadata and controls

56 lines (40 loc) · 1.6 KB

An interactive way to explore 2d arrays (especially for correlations and coincidences).

Click here to open it!

Correlation Explorer - Screenshot

Intro

You can:

  • Load any csv file with numerical values (first column and first row specify the labels).
  • Lookup values.
  • Sort rows and columns.
  • Explore multidimensional data (e.g. 100 dimensions).

Examples of usage:

  • Exploring correlation matrix.
  • Looking at projections of variables on principal components.
  • Discovering dependencies between two sets of variables (e.g. countries vs diseases).

License

CC-BY Piotr Migdał at Compass Inc.

Loose notes

To do

  • examples
  • polishing sorting
  • autodetect of correlation vs count data

Feature creep

  • ? Gray lines when mouse-on
  • Importance, principal component, linkage sorts?
  • Make sorting more consistent
    • Alphabetic and Orginal as buttons
  • pseudo-symmetric sort, when first one axis, then automatically along the other
  • Bootstrap or some other nice interface? Or rather keep it bare-bone?
  • Easier search for / filtering variables?
  • Some listing of previously loaded csv files?
  • Some scales, not only -1 to 1 (so it will be usable also for counts)

Ideas

  • How to work with many variables?
    • Make first n variables bigger?
  • If and how allow more datapoints per pixes (e.g. also Spearman correl)
  • How to expand it into a 2d scatter/histogram plot per each correlation?
  • Show how much position changed?