Criteria for an appropriate sorption model based on statistical analysis

Moisture sorption behaviour of food systems is valuable in analysing and predicting their dehydration characteristics. Many sorption models are in use and selected ones were listed. An appropriate model is recommended on the basis of its statistical assessment. Regression parameters (coefficient of determination, standard error and mean relative deviation modulus) from about 290 data sets were presented and related one to another. A low mean relative deviation modulus did not result from either a high coefficient of determination or a low standard error of estimate in all the cases. The residuals from the Oswin, Smith, Henderson., GAB, ChungPfost, and Halsey models on typical sorption data were discussed. No one parameter was found to solely indicate a perfect fit, and sorption models will be better assessed on more than one statistical parameter before being chosen.