Is there ANY exploitable structure, or is this white noise? Ljung-Box portmanteau at multiple lags, Lo-MacKinlay variance ratios (heteroskedasticity-robust) at q=2/5/10, runs test, and a vol-clustering check on absolute returns — fused into a strategy-archetype verdict per series, momentum-flavored
Is there ANY exploitable structure, or is this white noise? Ljung-Box portmanteau at multiple lags, Lo-MacKinlay variance ratios (heteroskedasticity-robust) at q=2/5/10, runs test, and a vol-clustering check on absolute returns — fused into a strategy-archetype verdict per series, momentum-flavored (VR>1), mean-reversion-flavored (VR<1), or nothing-in-the-mean. Use before fitting ANY model, for "i