Principal Component Analysis (PCA)

https://twitter.com/allison_horst/status/1288904459490213888

PCA seeks to represent multiple variables that covary, with a smaller number of new predictors. This is done by seeking weighted linear combinations of the original set, that “explains” most of the variability of the full dataset.

The result of a PCA, is these weights (loadings), which are instructions on how to transform the original variables into principal components.

To interpret principal components, we use a chart that plots PC loadings vs original variables, and screeplots.