Principal Component Analysis

Dimension Reduction Techniques applied in R

Dimension Reduction techniques

This practical is divided into two parts.

First I show you how to fit PCA and factor analysis models in R.

Second, we do a practical application on currency data - fitting a linear regression using the components derived from the reduced dimension.

• Practical 4

• Practical 4 Bonus

Class Video

• Video Link

DIY Section

Part I

Run a PCA on active manager returns in SA.

First update fmxdat to get latest active manager returns:

devtools::install_git("Nicktz/fmxdat")

Active_Managers <- fmxdat::ASISA
Local_Benchmarks <- fmxdat::Local_Indexes

Now - use the practical to do some analytics on which active managers provide useful differentiations. Also, calculate the sum of the first three principal components’ eigenvalues, on a rolling 60 month basis.

Part II

Create a story for a blended fund of funds strategy.

• Invest in the past 3 years’ 10 top performing strategies on an equal weighted basis at the beginning of each year.

• Create a plot similar to the one below to show the performance vs active peers of your blended strategy:

Example

• Tip: use geom_tile.
• Create a flexible function that allows you to view this on a past 1, 3, 5 and 10 year basis.