Both biological and environmental data are needed to compute tolerance values. For each sampling station a list of taxa should be developed recording one of the following: presence/ absence; abundance; and/or relative abundance of each individual taxon. A large sample size is preferred because in presence/absence models at least 10 occurrences of each taxon is required for each degree of freedom in the statistical model. For example, a typical model relating a taxon's occurrences with a single environmental gradient requires three degrees of freedom, so the data set would have to include at least 30 sites where the taxon was observed, and at least 30 sites where the taxon was absent.
Each biological sample should have a matched environmental dataset collected at the same time and location that includes all of the environmental values of interest, such as pH or stream temperature. A broadly sampled gradient of the environmental variables is crucial to developing tolerance values. A good sampling design should include collecting data near both ends of the environmental gradient so that the biological response to the gradient can be observed.
In general, the higher the taxonomic resolution used for tolerance values, the better for diagnosing environmental stressors. Tolerance values developed for periphyton generally require species level taxonomy or higher. Although family level taxonomy has been used to develop tolerance values for macroinvertebrates, genus (or species) level is preferable.