Experimental Design
An empirical result is credible when its hypothesis, information boundary, comparison, sampling procedure, uncertainty, and failure conditions are explicit before interpreting performance.
Central questions
- What observation would contradict the hypothesis?
- Which data selected the model and which evaluated it?
- Are dependence, multiple comparisons, and regime changes addressed?
Planned model
Run the same signal through random splits, chronological splits, purged evaluation, and repeated parameter search to expose leakage and selection bias.