Search

The sampling, subsampling (both coarse and fine ground meal), and analytical variances associated with testing shelled corn for aflatoxin were estimated by the use of 500 g samples, 50 g subsamples, and the CB method of analysis. The magnitudes of the variance components increased with an increase in the aflatoxin concentration. Functional relationships were developed to predict the variance for a given aflatoxin concentration and any size sample, subsample, and number of analyses. At 20 ppb total aflatoxin, the coefficient of variantion associated with a 4.54 kg sample, 1 kg subsample of coarsely ground meal (passes a #14 screen), a 50 g subsample of finely ground meal (passes a #20 screen) and one analysis were 21, 8, 11, and 26%, respectively.

Mean and the consensus scores from flavor and the texture profile analyses of selected food products were compared to determine which is superior for the purpose of statistical analysis. Results from the analysis of variance indicated that the mean scores gave a smaller coefficient of variation than the consensus scores for most of the character notes evaluated. When principal component analysis was applied to both the mean and the consensus scores, mean scores produced a higher “cumulative proportion of variance accounted for” by the component. The results of this study suggest that using the mean scores from flavor profile or texture profile analyses is superior to using consensus scores.

A flatoxin has been confirmed to be an important car-cinogen in certain animals and its contamination has been discovered in such agricultural products as corn, barley, rice, soybeans, wheat, cowpeas, sweet potatoes, cotton seed, and various types of nuts from all over the world (Anon., 1974; Anon., 1976). Detoxification by some method is essential if the hazards of consumption are to be removed. Automatic removal of the con-taminated portion of a product would be an effective means of reducing the hazards. There are several studies on record of the correlation between fluorescence and aflatoxin contaminated nuts and seeds (Ashworth and McMeans, 1966), white corn (Fennel et al., 1973), and pecans (Tyson and Clark, 1974). Bothast et al. (1975) showed the relationship of bright-greenish-yellow (BGY) fluorescence emission of wheat, oats, barley, sorghum and corn to aflatoxin con-tamination. Dickens and Welty (1975) observed that an average of 7 percent of the shells of the nuts in samples from 46 naturally contaminated lots of pistachio nuts emitted a BGY fluorescence, and this portion of the nuts contained an average of 50 percent of the aflatoxin in the lots. Farsaie et al., (1978) while studying the fluorescence of pistachio nuts under long wavelength ultra-violet light (LW-UV) observed that the nuts could be classified into four basic categories: bright-greening yellow (BGY), bright yellow (BY), normal (N), and purple (P), fluores-cence. Several sorting indices were developed and tested (Farsaie etal., 1978). McClure and Farsaie (1977) developed a fiber optic photometer for making fluorescence measurements at two wavelengths simultaneously. Work with this photo-meter indicated that the log ratio log10 (I490/I420) was the best optical index for dividing pistachio nuts into the four fluorescence categories (BGY, BY, N, and P). This paper describes the design and performance of an automatic sorting machine which employs this index.

The sampling, subsampling, and analytical variance associated with testing cottonseed for aflatoxin were estimated by use of 4.54 kg samples, 100 g subsamples, and the Velasco method of analysis. Regression analysis indicated that each of the above variance components is a function of the concentration of aflatoxin in the populations tested. Functional relationships are presented to determine the sampling, subsampling, and analytical variance for any size sample, subsample, and number of analyses.

Flavor profile data for beefsteaks, fish gels, frankfurters, peanuts, peanut butters, and baked sweet potatoes were subjected to principal component analysis to study the interdependency and the underlying dimensions of the sensory flavor characteristics in order to simplify the method of flavor profile analysis. The character notes used by the panel to describe the flavor profile of each food were not completely independent. The underlying dimensional structure had a number of dimensions equal to approximately 1/3 the number of character notes which explained about 3/4 of the total variance. Flavor profile analysis for each of the evaluated foods could thereby be simplified through reducing the number of character notes to about 1/3 the original number.

Sensory texture profile data for butters, cheeses, fish gels, frankfurters, peanut butters, raw potatoes, and baked sweet potatoes were subjected to principal component analysis to study the interdependence and the underlying dimensions of the sensory textural characteristics. The results indicated that the character notes used by the panel to describe the texture profile of each product were not entirely independent but formed a pattern of interdependency in the underlying dimensional structure. The texture profile could be described by about ½ the orthogonal factors (underlying dimensions) compared to the original character notes and still explain about 72% of the original variance. It is concluded that the texture profile analysis for each of the products in this study could be simplified by reducing the character notes being evaluated.

A subsampling mill was constructed with 19 subsample spouts similar to the 1 or 2 subsample spouts on the conventional subsampling mills. Samples of peanuts, each of which contained 1 kernel made radioactive by neutron activation, were comminuted in the mill. For treatment M the discharge from each spout was kept separate. For treatment R, the discharge from all 19 spouts was blended together in a twin-shelled blender and subdivided with a riffle divider into 16 subsamples which averaged the same weight as the M subsamples. Radioactivity counts/gram of each subsample were measured. Eleven samples were comminuted for each treatment which produced a total of 209 M samples and 176 R samples. An analysis of variance on the pooled data from each treatment showed a mean square error of 1318.3 for the M treatment and 1278.6 for the R treatment. The null hypothesis that the mean square errors for the two treatments are equal was not rejected by the F test (P=0.421).

Suitability of the negative binomial distribution for use in estimating the probabilities associated with sampling lots of shelled peanuts for aflatoxin analysis has been studied. Large samples, called “minilots,” were drawn from 29 lots of shelled peanuts contaminated with aflatoxin. These minilots were subdivided into ca. 12 lb samples which were analyzed for aflatoxin. The mean and variance of these aflatoxin determinations for each minilot were determined. The shape parameterk and the mean aflatoxin concentrationm were estimated for each minilot. A regression analysis indicated the functional relationship betweenk andm to be:k=(2.0866+2.3898m) × 10−6. The observed distribution of sample concentrations from each of the 29 minilots was compared to the negative binomial distribution by means of the Kolmogorov-Smirnov test. The null hypothesis that each of the true unknown distribution functions was negative binomial was not rejected at the 5% significance level for all 29 comparisons.