Implication of correlations among some common stability statistics - a Monte Carlo simulations

Area of Science:

• Agricultural science
• Biometrics
• Statistical genetics

Background:

• Stability analysis in multilocation trials commonly employs a mixed two-way model.
• Environmental variance (S i (2)) and ecovalence (W i) are frequently used stability measures.
• Theoretical expectations under the two-way model suggest identical rank orders for S i (2) and W i.

Purpose of the Study:

• To investigate the discrepancy between theoretical rank orders and low empirical rank correlations of stability measures.
• To determine if the mixed two-way model is appropriate for real-world multilocation trial data.
• To assess the impact of sampling errors on the observed low rank correlations.

Main Methods:

• Monte Carlo simulation was utilized to model the behavior of stability statistics.
• The study analyzed the relationship between environmental variance (S i (2)) and ecovalence (W i) under sampling variations.
• Statistical analysis focused on rank correlations and expected values of stability measures.

Main Results:

• Monte Carlo simulations revealed that sampling errors are the primary cause of low empirical rank correlations between S i (2) and W i.
• The observed low rank correlations do not invalidate the appropriateness of the two-way mixed model.
• The study confirmed that theoretical rank orders of S i (2) and W i are consistent with the two-way model.

Conclusions:

• The mixed two-way model remains a valid framework for multilocation trial stability analysis.
• Low empirical rank correlations between S i (2) and W i are attributable to sampling variability.
• Further discussion includes homogeneity tests for S i (2) and the model's implications for classifying stability statistics.