PCA explorer
Principal Component Analysis squeezes a table with many numeric columns into a couple of dimensions you can actually look at. In the R world, the reference tool for this is Factoshiny, the Shiny interface of the FactoMineR package.
This post is an attempt to rebuild that experience with React and d3.js only. Pick a dataset, toggle variables, and read the 2 factor maps. No R session, no server, everything runs in your browser.
Dataset
50 individuals · 4 active variables
Active variables
Display
Axes
Individuals factor map
Individuals close to each other are alike
Variables factor map
Arrows close together are correlated variables
Scree plot
How much information each dimension carries
Eigenvalues
| Dim | eigenvalue | % var | cumul. % |
|---|---|---|---|
| 1 | 2.480 | 62.0 | 62.0 |
| 2 | 0.990 | 24.7 | 86.8 |
| 3 | 0.357 | 8.9 | 95.7 |
| 4 | 0.173 | 4.3 | 100.0 |
Variable contributions (%)
■ Dim 1 ■ Dim 2
Contact
👋 Hey, I'm Yan and I'm currently working on this project!
Feedback is welcome ❤️. You can fill an issue on Github, drop me a message on LinkedIn, or even send me an email pasting yan.holtz.data with gmail.com. You can also subscribe to the newsletter to know when I publish more content!
