In relation to my previous question (Who influences Forex prices and by how much?) I have an raw idea how to determine how much is Forex influenced externally and how much is its behavior given by its previous states (reaction of Forex trades on Forex prices).
The idea is based on relative comparison with totally random signal, and completely predictable signal via multiple predictive models. The raw idea:
- the random signal should be something strongly similar to Forex time series (levy flight?)
- the predictable signal could be something like sum of a huge number of harmonic waves (similar FFT spectrum like the Forex time series, if possible)
- as a predictive models I would use many options ranging from auto-regressive models to deep learning models.
The outcome should be some relative scale how much (on every granularity) is the Forex time series close to the random signal or the predictable signal in the terms of predictability. This outcome should help me to understand, how much is Forex influenced from outside vs inside.
Is this experiment idea valid? Or is completely wrong? If is it valid, what potential problems I can encounter? What kind of time series (random and predictable) I should use?