nx_ice_stabiliser.nx

nx_ice_stabiliser.nx -- THE HYDROCOLLOID TERM, and the reason it is much smaller than this lane assumed. ===== WHY THIS EXISTS ============================================ nx_ice_recrystal models Ostwald ripening and nx_ice_distribution derives the geometry, but neither knows what a stabiliser does. The gap ledger named this four separate times, each time in the same words: "no stabiliser/hydrocolloid term, and locust bean gum is exactly how industry slows this mechanism". **THAT SENTENCE CONTAINS TWO ERRORS AND THIS ORGAN CORRECTS BOTH.** ERROR 1 -- THE REFERENCE PRODUCT DOES NOT USE LBG. Ben & Jerry's Vanilla declares GUAR (primary) and CARRAGEENAN (secondary). Haagen-Dazs Vanilla declares no stabiliser at all. The lane had been planning to model the competitor's lever using a gum the competitor does not buy. ERROR 2 -- AND THE LARGER ONE: **IN FINISHED ICE CREAM THE EFFECT IS NOT STATISTICALLY SIGNIFICANT.** Flores & Goff 1999 (JDS 82:1408-1415) cycled full ice cream at 0.2% stabiliser and every stabilised sample shared significance letter "a" with the unstabilised control. The mean X50 values LOOK like a 15-25% improvement. The study did not find one. ★★★A DIFFERENCE VISIBLE IN A TABLE IS NOT A DIFFERENCE THE STUDY FOUND. This organ therefore stores the Flores means AND the significance verdict together, and stab_flores_difference_is_significant() returns 0 as a FUNCTION, so no downstream planner can read the means alone and price a lever that the evidence does not support. ===== WHERE THE EFFECT IS REAL =================================== In MODEL SOLUTIONS the effect is real, and it is CONDITIONAL ON MILK PROTEIN rather than monotonic in dose. Regand & Goff (Guelph thesis Table 5.2), 5 cycles -3.5/-6 C, recrystallisation rate in um per cycle: no milk solids with milk solids none 4.02 4.59 LBG 4.05 3.17 carrageenan 3.80 3.02