The measure and the gaze: data analysis in agricultural sciences

Abstract

A 30-minute talk that sets out three mirrors of data analysis in agricultural and animal production sciences, with Goodhart’s law as the common thread. The first uses SCImago Journal & Country Rank data to describe Colombian scientific output in agricultural and biological sciences: more published documents, fewer citations per document, national self-citation close to 17 % and journals concentrated in quartiles Q3 and Q4. The second reviews the usual statistical toolkit (ANOVA, mean comparisons, p-values) in light of the 2016 ASA statement and proposes broadening it with generalized additive models, mixed and Bayesian models, conformal intervals, simulation and machine learning. The third examines agentic AI systems, their jagged intelligence and the need for audit cycles and human oversight: research quality depends on the judgment with which tools are used.