Using Autocorrelation to Analyze Trading Time Series
Summary
The document introduces autocorrelation as the relationship between observations in a time series and their lagged values. It explains that positive autocorrelation can indicate persistence, while negative autocorrelation can suggest reversal, and describes calculating autocovariance and normalizing it by variance to obtain coefficients between minus one and one. An autocorrelation function plot can show how these relationships change across lags. It also distinguishes ACF, which includes indirect lag effects, from partial autocorrelation, which isolates direct effects.
A practical example outlines calculating rolling autocorrelation on Apple closing prices and generating threshold-based long and short signals, then plotting them. The article presents trend-following and mean-reversion interpretations as possible uses, not reliable forecasts. It notes that autocorrelation can create modeling problems and false signals, and that residual autocorrelation can undermine regression inference. The excerpt does not provide complete strategy results or establish profitability; signal behavior depends on data choices, lag, threshold, and market conditions.
Key ideas
- Autocorrelation measures the relationship between a time-series observation and its lagged values.
- Positive autocorrelation can suggest persistence, while negative autocorrelation may be consistent with reversal.
- ACF includes direct and indirect lag relationships, while PACF focuses on direct relationships after accounting for intermediate lags.
- Rolling autocorrelation and threshold rules can be used to illustrate trading signals, but do not guarantee predictive value.
- Autocorrelation can violate model assumptions and make statistical inference or trading signals unreliable.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.