Analysis of the fluctuating microbial counts in commercial raw milk--a case study

Microbial counts of raw milk from eight commercial dairy farms recorded over a period of about 2 years were obtained from a processing plant. All the counts formed irregular fluctuating series whose statistical properties were examined. In all eight cases the autocorrelation function indicated that the counts were random and had no underlying periodicity. This suggested that the fluctuations, at least to some extent, reflected the combined effect of numerous factors, some unknown, and that these factors operate randomly and independently. In five out of the eight cases the fluctuating pattern of the record could be described by a probabilistic model based on the assumption that the counts have a lognormal distribution. The validity of the model was confirmed by comparing the calculated frequencies of counts exceeding preselected levels with those actually observed in fresh data. The model enabled ranking the five sources on the basis of anticipated frequencies of excessive counts rather than on the basis of mean performance. Two sequences had clearly identified regions with different statistical characteristics. They were interpreted as a reflection of changes in the sanitary conditions at the corresponding farms. In one case, both the overall magnitudes and the fluctuation amplitude of the counts progressively decreased, a situation for which the described model as formulated was inapplicable.