# Accreditation of new technologies for predicting intramuscular fat percentage: Combining Bayesian models and industry rules for transparent decisions

## TL;DR

This paper introduces a new method for evaluating the accuracy of technologies that measure fat in meat, using fewer samples than traditional methods.

## Contribution

A novel statistical method combining Bayesian models and industry rules for efficient and transparent accreditation of IMF% prediction technologies.

## Key findings

- The new method required only 25 samples per IMF% range quarter for 80% likelihood of passing accreditation.
- The traditional rules-based approach needed at least 200 samples per quarter with only 50% likelihood of passing.
- An online App was developed for commercial users to test their technologies' accuracy.

## Abstract

The experiment evaluated a method for statistically assessing the accuracy of technologies that measure intramuscular fat percentage (IMF%), enabling referencing against accreditation accuracy thresholds. To compare this method to the existing rules-based industry standard we simulated data for 4 separate devices that predicted IMF% across a range between 0.5–9.5% for sheep meat. These devices were simulated to reflect increasingly inaccurate predictions, and the two methods for statistically assessing accuracy were then applied. We found that for the technology which only just meets the accreditation accuracy standards, as few as 25 samples were required within each quarter of the IMF% range to achieve 80% likelihood of passing accreditation. In contrast, using the rules based approach at least 200 samples were required within each quarter of the IMF% range, and this increased the likelihood of passing to only 50%. This method has been developed into an on-line analysis App, which commercial users can freely access to test the accuracy of their technologies.