Tetsuo Aishima

Flavour evaluation by multiple regression analysis of gas chromatographic patterns in soy sauce

Abstract The relationship between the gas chromatographic patterns and sensory tests on soy sauce aroma was investigated by modified multiple regression analysis. As a result good linear relationship between the two was observed on two different sample sets, with multiple correlation values of 0·949 and 0·966, respectively. This means that the result of the sensory test on a sample can be estimated from the gas chromatographic pattern. Next, the contributing proportions were calculated on all the peaks of a gas chromatographic pattern. These values mean the quantification of the importance of each peak for the whole aroma. The calculation of the contributing proportion leads to the identification and evaluation of the components responsible for the whole aroma and accordingly the technique obtained here offers a first step toward general studies of food flavour.

Objective specification of food flavour. Analysis of gas chromatographic profiles of soy sauce flavour by stepwise regression analysis

Abstract The relationship between gas chromatographic profiles and sensory tests, which were carried out by an ordering method, was investigated by stepwise regression analysis. The, results indicated that sensory qualities were linearly related to gas chromatographic profiles. The multiple correlation coefficient (R) increased with the increase of step number, and exceeded 0·9 at step 10. R reached 0·968 at the last step number, 43. On the other hand, the standard error of estimate reached a minimum value at step 28 and then began to increase gradually. The most predictive regression model for sensory test panel acceptability was calculated for each step and the resulting calculated models were tested by substituting the gas chromatographic data. The results showed good accuracy for the estimation of sensory quality.