The FactorAnalytics project is an open source package containing fitting and analysis methods for the three main types of factor models commonly used in conjunction with portfolio construction, optimization and risk management. These include:
- fundamental factor models
- time series factor models
- statistical factor models
The purpose of this project is to provide basic features and capabilities close to those of commercial portfolio optimization and risk management products.
In addition, the package contains functions for related risk and performance attribution estimates including:
- volatility
- VaR
- ES
- factor-contributed vs idiosyncratic returns
- factor model Monte Carlo
- multiple imputation methods for simulating returns and backfilling unequal histories
- functions to create tabular displays of risk and performance reports
Install from source is the preferred method of installation. To do so, your local git installation will need to support git large file storage (LFS). Instruction for installing LFS can be found HERE
Plenty of documentation exists for cloning github repos and building R packages locally. A simple search specifying your IDE and OS should do.
Note a bug in remotes/devtools prevents the package from being loaded using install_github("braverock/FactorAnalytics"). Please see r-lib/remotes/issues/637.
S&P Global Market Intelligence has kindly provided firm fundamentals data
referred to as “scores” or “alpha factors” for educational use in the open source FactorAnalytics R package. The data is contained in the R data frame object “factorDataSPGMI” consisting of the following cross-section of scores for approximately 300 stocks from 1990-2015: AccrualRatioCF, AnnVol12M, Beta60M, BP, Chg1YEPS, DivP, EBITDAEV, EP, EQ-style,
LogMktCap, PM12M1M, ROE. This data greatly facilitates the educational value to users of the fundamental factor model in FactorAnalytics. The package developers wish to thank S&P Global Market Intelligence for kindly contributing this data, which will improving learning outcomes for all.
From the first paragraph
The overarching long-term goal of the fundamental factor model (Ffm) development in the FactorAnalytics packages is to replicate a large proportion of the non-proprietary models and model fitting and analysis methodology that is contained in commercial portfolio construction and risk management products such as MSCI Barra, Axioma, Northfield, etc. Furthermore our goals include the implementation cutting edge methods to support portfolio construction and risk management that are not much available in commercial products, such as global optimization, unequal histories and other missing data handling, highly robust covariance matrix estimators and their application to multivariate exposures and returns outliers, optimal bias robust regression, factor model Monte Carlo, new methods for handling serial correlation that improve upon traditional HAC methods, etc.
R Script and slides used in Prof. Douglas Martin's "Fundamental Factor Models in FactorAnalytics" Pre-Conference Seminar.
Click here for the background slide deck for the Boston useR group talk by Prof. Doug Martin.