Posted at 23 November 2018
Classical methods to assess pet food palatability performance include preference trials conducted with dogs and cats expert, semi-expert or in-home panels. A huge amount of information such as animal daily consumption data can be collected from these tests, nevertheless it recurrently happens that some data are not available and/or reliable due to veterinary cares or unusual animal behaviors during the trials. The study conducted by Diana Pet Food's statisticiens aimed at evaluating the efficiency of seven completion methods to recover cat consumption missing data in the particular case of pet food external preference mapping.
In the case studied, ImputePCA, missforest and kNN were the most relevant methods to recover the palatability missing data. They allowed to successfully refill missing values and to perform an external preference mapping with the initial size of the database.
These results opens up new perspectives for pet food palatability assessment. Indeed, completion methods could be interestingly tested to recover various other types of palatability missing data obtained from classical two-bowls palatability test, kinetics of consumption…
Results presented at Missdata 2015
Authors: J Roguès @Diana Pet Food, Elven, France
Want to know more about this poster? Contact us here