How long should a signal remember?
A proposed comparison of short and long memory windows using the same input signal.
The question
Does a longer memory window retain useful information, or mostly delay a response? The study would compare a small set of predeclared windows while holding the raw input and prediction target fixed.
Study status: proposed, not run. This sample note exists to review the reading experience and research layout. It does not report an observed market result.
Proposed test
Choose one existing measurement and record when each observation becomes available. Define the smoothing rules before running the comparison. Use chronological evaluation periods and keep the same observations across all variants.
Study record
| Item | Current state |
|---|---|
| Raw input | To be selected and documented |
| Memory windows | Predeclare before analysis |
| Evaluation period | Not selected |
| Measured result | Not run |
What could mislead us
A smoother-looking chart would not be evidence of a better signal. The comparison would need to separate stability, response time, and any predictive improvement.
Results
No results are available. The finished version would show the sample size, excluded observations, comparison with the baseline, uncertainty, and links to reproducible artifacts. Empty result fields are intentional.
Next step
Measure response delay as well as the chosen prediction metric. Publish the comparison only after the dataset, code revision, and output artifacts are available.
Return to the research notebook or explore the Memory series.