A regression-based statistical framework for continuous nutritional inputs in poultry nutrition studies: beyond ANOVA

Nutritional studies in animal science frequently evaluate graded dietary inclusion levels using one-way ANOVA followed by multiple mean comparison procedures, despite the fact that dietary inputs are inherently continuous variables rather than independent categorical treatments. This widespread analytical practice limits biological interpretation, reduces statistical power, obscures dose-response dynamics, and prevents estimation of optimal nutrient inclusion levels. The present study proposes a standardized regression-based analytical protocol for animal nutritional experiments involving continuous dietary inputs and demonstrates its application using published broiler response data obtained from graded supplementation of herbal mixture powder (HMP; 0, 1, 2, and 3 g/kg diet). Mean response data from growth performance, carcass traits, blood biochemistry, antioxidant status, immune response, meat quality, and cecal microbiota were re-analyzed using linear and quadratic regression models. Standardized regression coefficients were additionally calculated to quantify relative effect sizes and rank biological trait sensitivity to dietary supplementation. Regression analysis successfully characterized monotonic, plateau, and curvilinear dose-response relationships that were not adequately captured by conventional mean separation procedures. Quadratic modeling enabled estimation of biological optimum inclusion levels, whereas standardized coefficients identified the most responsive physiological indicators independent of measurement scale. Growth traits, lipid metabolism variables, antioxidant biomarkers, and immunological parameters exhibited the greatest responsiveness to HMP supplementation, while feed intake and intestinal traits demonstrated comparatively lower sensitivity. The proposed analytical framework provides predictive capability, quantitative biological interpretation, optimization potential, and cross-trait sensitivity evaluation that cannot be achieved using ANOVA-based mean comparison alone. These findings demonstrate that regression analysis complements conventional ANOVA in studies involving graded dietary inputs by quantifying dose-response relationships, estimating biologically relevant optima, and providing predictive interpretation beyond overall treatment comparisons. The present work provides a practical statistical protocol for improving the analytical rigor, biological relevance, and reproducibility of animal nutrition research involving graded dietary inputs.