ABSTRACT

V. D. Fedorov, V. N. Maximov and V. G. Bogorov obtained the data from an experiment on the composition of a nutritive medium for green sulphur bacteria chlorobrium thiosulphatophilum. The bacteria were grown under constant illumination at a temperature of 25–30 °C: the yield was determined during the stationary phase of growth. Although these data in fact arose from an experiment, the haphazard character of the design means that the analysis is in many ways more typical of that of unbalanced data arising in observational studies. The availability of an independent external estimate of standard deviation is in the case very important. While there is no uniquely optimal way of approaching the analysis, the following is sensible. Fit a model containing only main effects, thus involving 11 parameters. The approach is thus to fit a small number of main effects which the first analysis shows to be appreciable, plus the corresponding interaction terms.