This repository contains experiments on the estimation of time-dependent sensitivities / factor loadings in factor models. Random factor models with time-dependent sensitivities are created and the performance of various estimation models are compared. The models implemented include
- Constant OLS linear regression
- Rolling OLS linear regression
- Exponentially weighted OLS linear regression
- Sensitivities modeled as random walk, estimated with a Kalman filter (henceforth "Kalman")
- Sensitivities modeled as a local trend model, estimated with a Kalman filter (henceforth "Kalman local trend").
Check the slides "Slides-TimeDependentEstimation.pdf" for more information.
Inspired by the paper: Bentz: Quantitative Equity Investment Management with Time-Varying Factor Sensitivities, in Dunis, Laws, Naim: Applied quantitative methods for trading and investment.