Stability Analysis in Crop Trials
Stability analysis quantifies how a genotype performs across environments. A stable genotype performs consistently; an unstable genotype is sensitive to environment. Several frameworks exist; Eberhart and Russell (1966) is the classical one and is what most agronomy curricula and many ICAR manuals teach.
Eberhart Russell parameters
The Eberhart-Russell model regresses each genotype's performance on the environmental index (the deviation of the environment mean from the grand mean). It produces three parameters per genotype: the mean (overall performance), the regression coefficient b (sensitivity to environment), and the deviation from regression S squared d (unpredictable instability). A stable genotype has high mean, b close to 1 and S squared d close to 0. A genotype with high mean and b less than 1 is well adapted to poor environments. A genotype with high mean and b greater than 1 is well adapted to good environments.
Reading the output
The output is a table with one row per genotype and columns for the mean, b, S squared d, and significance tests for b versus 1 and S squared d versus 0. Two F tests matter: F_b tests whether the regression of genotype on environment is significant; F_S squared d tests whether deviations from regression exceed pooled error. A genotype is stable in the strict Eberhart-Russell sense if both b is close to 1 and S squared d is not significant.
When to look beyond Eberhart-Russell
Eberhart-Russell is a univariate, regression-based model. It does not capture multi-pattern interactions. AMMI (Additive Main effects and Multiplicative Interaction, Gauch 1988) and GGE biplot (Yan and Kang 2003) give a multivariate view of genotype-by-environment interaction. Use those when the regression assumption is too restrictive or you want a visual mega-environment analysis. StatVeda exposes the Eberhart-Russell model directly; AMMI and GGE engines exist in src/lib/stats/ammi-gge.ts but are not yet wired into the UI.