Comparison of Item Response Theory Models for Data Fit of Quality Control Measurements for Maillard Reaction Browning in Foods
Date of Award
Doctor of Philosophy
Educational Leadership, Research and Technology
Dr. Brooks Applegate
Dr. Jessaca Spybrook
Dr. P. Doug Williams
Maillard reaction, Hem Response Theory, Browning, Qualitv control, sensory, polytotomous
In foods, browning is a complex set of reactions known as the Maillard Reaction (MR). As food is cooked, MR yields flavor, color and changes to the chemical properties of the initial food item. It is possible to undercook or overcook the food item, resulting in a less browned or burnt product. Item response theory (IRT) is often used to place subjects and instruments on the same continuum of the measured latent trait. IRT has been widely used in the fields of educational measurement and psychology, but few studies combine IRT and food science. This research tested IRT modeling for the latent trait browning in foods to determine quality control parameters for the MR.
Samples representing the anticipated range of browning in a production setting were created using a laboratory simulation. An array of test items measured the MR, including analytically measured physical properties of the food and stimulus based sensory human responses to the food. A comparison of item response methods was conducted (Hyperbolic Cosine [HC]; Graded Response Model [GRM], and 2-Parameter Logistic [2-PL]). Data analysis included a comparison of model fit statistics. Investigation into monotonicity of items was conducted.
IRT was adequate to model browning in foods, as tested using chi-square and standard error measurements of the model. Identified IRT parameters were used to set predict the degree of browning in samples, then used to set specifications for the quality control of MR. Items included in the quality control test for MR were both binary and polytotomous.
This research enhances the knowledge base of IRT modeling for multidimensional traits that can be broken down into unidimensional traits. These statistical techniques can be used for ideal point modeling and to determine best model fit for polytotomous items. Recommendations for the future include implementing these new items as a systematic approach at predicting quality in future production batches of cooked foods.
Restricted to Campus until
Barrett, Polly, "Comparison of Item Response Theory Models for Data Fit of Quality Control Measurements for Maillard Reaction Browning in Foods" (2015). Dissertations. 513.