FLAVOR CHANGES IN WHOLE MILK POWDER DURING STORAGE

Abstract Multiple linear regression analysis was used to determine predictive equations relating the intensities of flavor properties in whole milk powder to volatile compounds. Flavor properties were analyzed by flavor profiling and volatile compounds by gas chromatographic headspace methods. Predictive equations with both linear and log‐transformed data of volatile compounds were determined and a selection was made from 25 quantified compounds. Predictive equations were also determined by selection from 15 flavor compounds; i.e. from compounds that occurred at concentrations near or above their odor threshold levels. A computer program was used which performed regressions for each flavor descriptor on all possible combinations of two or three volatile compounds. Several predictive equations with high correlation coefficients were determined for each flavor descriptor. For some descriptors it was possible to determine predictive equations using hexanal, an easily analyzed compound in these samples.