Prediction of fillet weight, fillet yield, and fillet fat for live river catfish (Pangasianodon hypophthalmus)

Abstract The objective of this study on river catfish ( Pangasianodon hypophthalmus ) was to predict fillet weight and fillet yield from body measurements on live fish, and likewise for fillet fat, from Distell Fish Fatmeter recordings. Fish at marketable size, from the breeding program at the Research Institute for Aquaculture No.2, Vietnam, were randomly sampled and recorded for fillet weight and fillet yield ( n  = 2767) and fillet fat ( n  = 50). For fillet weight and fillet yield, the following body measurements were used; body weight, standard length, and volume, together with length, height, width and circumference at four positions along the body. The fish were also filleted, and the fillet weight (g) and fillet yield (%) were recorded. For modelling of fillet fat, the average of three readings with the Distell Fish Fatmeter was done at nine positions on each side of the fish. These fish were then filleted, and fillets were chemically analyzed for fat content (%). For fillet weight and fillet yield, a random sample of 200 fish were used for estimation, while the remaining (2567) were used for testing the prediction power. For fillet fat, all relevant records were used for estimation while in cross validation one record was left out for prediction and the remaining were used for estimation. Multiple regression procedures with forward selection of variables were used throughout. The final prediction equations were those resulting in the least root mean squared error of prediction, with the correlation between predicted and observed values for fillet weight, fillet yield and fillet fat being 0.93 (5 variables), 0.86 (4 variables) and 0.85 (4 variables), respectively. However, due to the limited sample sizes used in the estimation, all prediction equations were biased. For fillet weight and fillet yield, the prediction equation is likely to be further improved by reducing measurement error in filleting. Additional explanatory variables should also be sought. For fillet fat, there is a need of evaluating numbers of recordings of the chemically analyzed fat and that of the Distell Fish Fatmeter. Increasing the number of records in the data used for estimation is expected to pick up additional marginal effects.