Statistical approach for comparison between methods of bacterial enumeration in drinking water

A general statistical procedure based on the likelihood ratio test is presented for the purpose of comparing estimates of mean bacterial density derived from different sets of data. This approach is much more appropriate than the conventional ways of analyzing bacteriological results (e.g., analysis of variance) which usually require previous transformation of the data. An illustrative application of the method compares three distinct titration techniques for enumerating heterotrophic bacteria in drinking water at 20°C incubation temperature. It was shown that both the standard plate count (SPC) and the membrane filter (MF) procedures supplied substantially the same information, whereas the microplate technique using the most probable number (MPN) for total bacterial enumeration could yield considerably different estimates: MPN values were significantly lower in three cases and significantly higher in one case out of a total of five experiments. The results consistently indicate a strong interaction between the technique used and the sample analyzed. Three different media (nutrient agar, R-2A low nutrient agar and m-SPC agar) were then evaluated for enumerating heterotrophic bacteria, using the MF technique at 48, 72 and 96 h of incubation time at 20°C. Although the media recovered approximately the same numbers of bacteria after 96 h of incubation, statistically significant discrepancies occurred after intermediate periods of incubation, perhaps because the relative rates of bacterial growth differed among media.