Detecting causal relations among symbolic time series.

Area of Science:

• Social Sciences
• Economics
• Time Series Analysis

Background:

• Symbolic time series analysis simplifies data by reducing noise.
• It aims to clarify the evolution of time series dynamics.

Purpose of the Study:

• To evaluate the reliability of causality tests on symbolic time series data.
• To identify potential failures and artifacts in these methods.

Main Methods:

• Application of established causality detection methods.
• Testing methods including transfer entropy, Granger's test, and Peter-Clark Momentary Conditional Independence (PCMCI).

Main Results:

• Causality tests on symbolic series may fail to detect true relationships.
• These methods can frequently generate spurious causal findings.
• Method performance varies with lag structure, alphabet size, and data characteristics.

Conclusions:

• Caution is advised when interpreting causality from symbolic time series.
• The choice of parameters significantly impacts the validity of causal inference.