Skip to main content
ARS Home » Northeast Area » Wyndmoor, Pennsylvania » Eastern Regional Research Center » Microbial and Chemical Food Safety » Research » Publications at this Location » Publication #380367

Research Project: Data Acquisition, Development of Predictive Models for Food Safety and their Associated Use in International Pathogen Modeling and Microbial Databases

Location: Microbial and Chemical Food Safety

Title: Acceptable prediction zones method for validation of predictive models for foodborne pathogens

Author
item Oscar, Thomas

Submitted to: Book Chapter
Publication Type: Book / Chapter
Publication Acceptance Date: 10/14/2021
Publication Date: 9/14/2023
Citation: Oscar, T.P. 2023. Acceptable prediction zones method for validation of predictive models for foodborne pathogens. In: Alvareng, V.O., editor. Basic Protocols in Predictive Food Microbiology. New York, NY: Humana Press. p. 185-210.

Interpretive Summary: Proper validation of predictive models for foodborne pathogens is important for two reasons. First, it provides model developers with an objective method to identify and repair prediction problems before models are provided to end users. Second, it provides end users with confidence that model predictions are reliable and can be used to make important food safety decisions. The focus of this chapter is to describe and demonstrate how the Acceptable Prediction Zones (APZ) method in the Validation Software Tool (ValT) can be used to properly validate predictive models for foodborne pathogens.

Technical Abstract: Proper validation of models is important because it provides users with confidence that model predictions are reliable. In addition, it helps modelers identify and repair prediction problems before models are provided to end users. This chapter describes the Acceptable Prediction Zones (APZ) method in the Validation Software Tool. The APZ method has criteria for test data, model performance, and model validation. These criteria ensure that comparisons of observed and predicted values are not confounded by differences in data collection and modeling methods. In addition, they ensure that model validation is accurate, unbiased, and objective. A tertiary model and secondary models (lag time, growth rate) for growth of Salmonella Typhimurium definitive phage type 104 (DT104) on chicken skin with native microflora are used to demonstrate the APZ method in the Validation Software Tool or ValT (vault).