Statistical methods in microbiology.

IM has discussed about the evaluation of novel microbiological methods with several professionals. His knowledge of statistical methods in microbiology bases on the lessons by Prof. Seppo Niemelä, who was (and still is) a well-known specialist in this not-so-well-known area of microbiology.

Testing of microbiological data is more complicated than similar analyses in chemistry. The main reason is the model of repeat distribution: chemistry follows the ordinary normal distribution but the colony counts of microbiological analyses are featured by the Poisson distribution. The reason for this difference is easy to understand: the count of molecules is overwhelming when compared to the limited count of colonies in microbiological cultivations. The dependence of variance on the mean of the data is another problem of colony count analyses, preventing the usage of parametric methods.

Luck enough, there are some non-parametric statistical analyses for Poisson-distributed data, helping the comparisons of means and trends of colony count results.

A novel problem seems to have arisen in microbiological evaluations since 70's. Many modern, automatized instruments of microbiology are not based on the measurements of chemical concentrations or counts of colonies but on specified metabolic activities of microbes. Examples of these analytical procedures are eg. measurements of impedance, turbidity, pH or CO2 production. Because these parameters are in a close connection to the growth rates - and to a new parameter, time - , their evaluations are very challenging procedures.

- More in next posts...