Applications of Stability Analysis for Single‐Site, Long‐Term Experiments

Abstract

AbstractStability analysis, which is the linear regression of treatment yield on the environmental mean yield, has been proposed for use in singlesite, long‐term experiments (treatments applied to the same plots year after year), without first determining if treatment yield reveals a significant relationship with year (experiment years plotted sequentially in time). This research was initiated to demonstrate potential problems associated with stability analysis on long‐term, wheat fertility experiments where treatment yield may be correlated with year. Three single‐site, long‐term experiments were evaluated using stability analysis and linear regression of treatment yield on year. Each study was a replicated, longterm, continuous winter wheat (Triticum aestivum L.) fertility experiment, receiving annual rates of N, P, and K fertilizer. Of the single‐site, long‐term trials evaluated, one was not suitable for stability analysis, as grain yield decreased significantly with each ensuing year that the experiment was conducted. In this case, a negative residual yield response was noted with time (year) in plots receiving N and no P. Therefore, the environmental mean in stability analysis no longer reflected the random influence of climate (e.g., low and high rainfall or temperature) which could be used to decipher variable treatment response in seasonably wet or dry years. Without prior evaluation of the relationship between treatment yield and year in single‐site, long‐term experiments, stability analysis can be inappropriate.