← Algorithms · Guinier analysis

Guinier analysis

Purpose. At low q, ln I vs q² is approximately linear for a compact particle. The slope gives the radius of gyration Rg; the intercept gives I(0). The chosen interval also defines where later tools (GNOM, BODIES) should start reading the curve.

What autosaxs does (custom) — adaptive Guinier. Rather than trusting a single fixed window or ATSAS AUTORG alone, autosaxs:

1. Enumerates many candidate Guinier intervals on the curve.
2. Fits each interval and validates the linear model on points beyond the fit window (validation R²), not only on the fitted points.
3. Selects a winner (preferring validation quality, then interval quality) and classifies the local shape: linear / upturn / downturn / chaotic.
4. Still reports comparison methods (fixed first-N points, AUTORG when available) for context; the adaptive result is the one used downstream when it succeeds.

In liveview you may override first / last (1-based point indices) for a fixed-interval fit; that suspends auto-processing until you resume.

How to read the UI. Quality / classification / Rg summarize the selected interval. A “chaotic” or poor class means the low-q region is unreliable for Rg even if a number is shown — treat GNOM and shape steps with caution.

Engines. Custom adaptive search (+ optional ATSAS AUTORG for comparison).

← Buffer subtraction · p(r) →