Sigmoids behaving badly: arXiv paperpost by Stuart_Armstrong · 2021-09-20T10:29:20.736Z · LW · GW · 1 comments
Building on a previous post on the unpredictability of sigmoid curves [LW · GW], we have a paper on the arXiv explaining why it is so hard to predict their inflection/turning points until after it has happened.
Sigmoids (AKA s-curves or logistic curves) are commonly used in a diverse spectrum of disciplines as models for time-varying phenomena showing initial acceleration followed by slowing: technology diffusion, cumulative cases of an epidemic, population growth towards a carrying capacity, etc. Existing work demonstrates that retrospective fit of data is often impressive. We show that in time series data, the future fit tends to be poor unless the data covers the entire range from before to after the inflection point. We discuss the theoretical reasons for this: the growth data provides little information about the damping term (and vice-versa). As a consequence, forecasting with sigmoids tends to be very unreliable. We suggest some practical approaches to improving the viability of forecasting sigmoid models.
Comments sorted by top scores.