Fiber–matrix interaction governs compressive strength in agave-bagasse-reinforced adobe: a factorial experiment with two-way ANOVA and competing mechanism analysis

Natural-fiber reinforcement is widely cited as a pathway to improve the mechanical performance of adobe, but reported effects on compressive strength are inconsistent across studies: some find improvement, others find degradation, and the choice of experimental conditions rarely disentangles the role of the fiber from that of the matrix. This study quantifies the coupling through a balanced factorial experiment. Forty-nine adobe specimens of 20 × 10 × 10 cm were manufactured with three granular compositions (sand-dominated, jal -dominated, and balanced, where jal is a regional non-plastic silt of Jalisco, Mexico) and four mass fractions of agave-bagasse fiber (0%, 0.5%, 1%, and 2%), and were tested under Mexican standard NMX-C-036-ONNCCE by an accredited external laboratory. Three complementary analytical tools are applied to the resulting dataset: (i) a two-way analysis of variance (ANOVA), (ii) a reinforcement efficiency index η with bootstrap confidence intervals, and (iii) a competing mechanism phenomenological descriptor f c ( ϕ ) = f c 0 + B ϕ e − ϕ / ϕ c − D ϕ that separates a saturating reinforcement term from a linear disruption term. The two-way ANOVA reveals a highly significant mixture–fiber interaction ( F 6,37 = 4.88 , p = 9.1 × 1 0 − 4 , and partial η p 2 = 0.44 ), which is stronger than either main effect and statistically demonstrates that the sign of the fiber effect is not an intrinsic property of the fiber but rather a property of the fiber–matrix pair. For sand-containing mixtures, the reinforcement efficiency index is η = 1.15 [M1, 95% bootstrap CI (0.96, 1.32)] and η = 1.22 [M3, (0.92, 1.59)] at the optimum ϕ * = 0.5 % ; a non-parametric bootstrap over 5,000 resamples places the optimum at ϕ * = 0.5 %