6 posts

Quant

Bodyweight as a time series: filters, lag, noise rejection and drawdown.

1. Can you use trading indicators on your weight? DEMA, HMA, KAMAThe price-only ones, yes. DEMA, TEMA, HMA and KAMA read one number per period, so nothing in them assumes a market. On one real log they finish 4.0 lb apart.
2. How long until your weight trend catches up? 28 days, or 5An EWMA at α = 0.1 needs 28 days to absorb 95% of a real change. DEMA gets there in 8 days and TEMA in 5 by subtracting an estimated lag, and both overshoot.
3. Why one trend method sits 4 lb above the scale (Kaufman KAMA)KAMA squares its smoothing constant, capping the alpha near 0.048. On 84 real weigh-ins it ignores 81.3% of daily movement and ends 4.0 lb above the last weigh-in.
4. Why your trend line follows every weigh-in (Hull moving average)Alan Hull's filter subtracts its own lag until the line leads the scale rather than trailing it. On 84 real weigh-ins it ignores 14.3% of daily movement.
5. Why the scale hasn't moved in 2 weeks: the signal-to-noise mathsTwo weeks is rarely long enough to tell a real change from ordinary scale noise. Work out your own noise-to-signal ratio, then how many days it takes to clear it.
6. Is it a real plateau? Why two weigh-ins a fortnight apart can't sayA stall is only a stall once it clears your own noise. On one user's log the two-morning read and the fitted slope disagreed on direction in 15 of 92 windows.