AbstractThe response measured in microbial toxicity tests is typically a continuous variable, and the associated error is an experimental error related to the physical measurement of the response and to handling differences between experimental units (e.g., test flasks). The traditional statistical theory for treating quantal (or categorical) data from tests with animals is based on the concept of a tolerance distribution for individual test subjects that is not appropriate for microbial tests. Data from microbial tests should be dealt with by conventional curve‐fitting techniques without any confusion with tolerance distributions. Using an empirical mathematical function that describes a sigmoid curve, concentration–response curves can be fitted by nonlinear regression directly onto the data or by weighted linear regression after a linear transformation of the data. Calculation of confidence intervals around the estimated EC figures (calculated effective concentrations that cause a certain percentage of inhibition; for example, EC10 for 10 inhibition) is less complicated with weighted linear regression on transformed data than it is with nonlinear regression. The paper provides a general discussion of the problem of analyzing data from microbial toxicity tests and gives some examples of feasible meth
展开▼
