Simulating ARMA Models and Building a Rolling Stock Strategy in R

Summary

This article shows how to simulate ARMA processes in R, inspect autocovariance and autocorrelation behavior, and use those models in a simple equity strategy. It describes using an ARIMA function from an R forecasting library for simulation and plotting autocorrelation and partial autocorrelation for autoregressive models. For the trading example, it fits ARMA models to Microsoft returns with an automated model-selection procedure, refitting each day on the history available up to that point. The strategy is long-only and uses a bounded search over autoregressive and moving-average orders, without seasonal terms, drift, or a mean.

Key ideas

• R’s ARIMA tools can simulate ARMA processes for exploring their behavior.
• ACF and PACF plots help inspect the dependence structure of autoregressive processes.
• The example refits an automatically selected ARMA model each day using prior stock returns.
• The author reports performance comparable to buy and hold and suggests varying the training span or testing short trades.
• The described strategy is a basic demonstration, and the article does not establish robust out-of-sample profitability.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.