The R software v4.0.3 (see R script) was used and the threshold of statistical significance was set at 0.05. The relationship between the dependent variables (weight, weight difference; see Experimental Data_weight) and the time or exposure pH was computed using linear or piecewise regression models that were compared using information criteria and verified for assumptions (see Graphical assumptions_Statistics_Experimental Data_Fig1C_upper panel; see Graphical assumptions_Statistics_Experimental Data_Left slope_Fig1C_lower panel). 

First, we used piecewise linear regressions of total body weight vs. exposure pH to estimate a tipping point for each day of measurement (see Statistics_Experimental Data_Fig1C_upper panel; see TippingPlot.0; see TippingPoint.0), defined as the value of the factor (pH) where the dependent variable tipped. We considered the first day of the ambient pH period as day 0 and therefore the experiment ran from day -14 to day 42. The significance of each slope was tested using Student’s t test. All regression models exhibited a pH tipping point above which slopes (factor > tipping point) were not significant. 

Thus, we assessed compensatory growth only below the tipping point (factor < tipping point) by computing weight differences across pH levels (g pH-1, dependent variable), given by the slopes, as a function of time using linear regression models (see Statistics_Experimental Data_Left slope_Fig1C_lower panel). We expected that (1) the weight differences below the tipping point would gradually fade after returning to ambient pH showing signs of compensatory growth (see Theory_reversible_data), or that (2) weight differences would increase indicating persistent growth stunting (see Theory_irreversible_data). Theoretical models corresponding to these hypotheses were constructed from the data of (Lutier et al., 2022).