Food sorption-isotherm dataset and Fit4 regression results for classical and reparameterised GAB models

This dataset contains experimentally measured food moisture sorption isotherms curated from the published scientific literature, together with curve-level nonlinear regression results generated using the Fit4 Excel fitting environment. The final curated dataset comprises 192 food sorption-isotherm curves. The curves represent a broad range of food materials and include both adsorption and desorption observations measured at different temperatures. Each retained curve contains at least four paired observations of water activity, awa_waw, and equilibrium moisture content, McM_cMc, and spans an awa_waw interval of at least 0.5. For each sorption curve, two regression formulations of the Guggenheim–Anderson–de Boer model were fitted to natural-log-transformed moisture-content data: Classical GAB parameterisation fitted to ln⁡(Mc)\ln(M_c)ln(Mc)The logarithm of the conventional GAB function was fitted using the original physical parameters XmX_mXm, CCC, and KKK. Reparameterised GAB formulation fitted to ln⁡(Mc)\ln(M_c)ln(Mc)The same underlying GAB model was fitted using transformed regression parameters designed to improve numerical conditioning while preserving the physical parameter domains after back-transformation. Both formulations were fitted under comparable nonlinear least-squares conditions using the Levenberg–Marquardt algorithm implemented in Fit4. The repository includes the observed sorption data, literature-source metadata, curve identifiers, fitting controls, fitted parameters, estimated parameter standard errors, regression diagnostics, goodness-of-fit statistics and error codes.