Skip to main content
ARS Home » Southeast Area » New Orleans, Louisiana » Southern Regional Research Center » Food Processing and Sensory Quality Research » Research » Publications at this Location » Publication #314621

Title: Statistical approaches to optimize detection of MIB off-flavor in aquaculture raised channel catfish

Author
item Zimba, Paul
item Grimm, Casey

Submitted to: Journal of Aquaculture Research and Development
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/16/2015
Publication Date: 2/15/2015
Citation: Zimba, P.V., Grimm, C.C. 2015. Statistical approaches to optimize detection of MIB off-flavor in aquaculture raised channel catfish. Journal of Aquaculture Research and Development. 6:3.

Interpretive Summary: Prior to harvest, pond raised catfish are sampled for off-flavor by professional flavor checkers working at the processing plants. Different plants have different procedures for sampling for off-flavor. Some plants sample two fish a week before harvest and some sample a single fish the day before harvest. This research determines the probability of detecting an off-flavor pond based upon the number of samples examined. Six fish is the minimum number of fish that should be sampled in a mixed population of on and off-flavor fish.

Technical Abstract: The catfish industry prides itself on preventing inadvertent sale of off-flavor fish. Typically, several fish are taste tested over several weeks before pond harvest to confirm good fish flavor quality. We collected several data sets of analytically measured off-flavor concentrations in catfish to assess the type of distribution (parametric/non-normal). Coincident measures of fat content were made on three subsections of each fillet. These data were then used to model the number of fish required to detect off-flavor in mixed populations containing on and off flavor fish. In fish collected from the same pond, off-flavor concentrations typically were not normally distributed, thereby, requiring specialized statistical procedures. Even with log transformation, data still violated assumptions of normality. We used a non-parametric approach, using fish samples that were ordered, and then randomly sampled 1000 times to determine the number of fish necessary to detect off-flavor. A sample of 40 fish was required to detect off-flavor when the population was nearly all on-flavor (97%) and <11 fish when populations contain >20% off-flavor fish. A sample size of six fish in a mixed population was effective in identifying off-flavor occurrence in 60% of having off-flavor present. Sampling more fish fewer times can more accurately identify ponds containing mixed flavor fish populations than the current sampling procedure.