Correlations and Taxonomic Level

We can also constrain the base set by using taxonomic information instead, or in addition to, geographic and temporal restrictions. For example, if we are studying the correlations between species we can take into base set only the locations in which there is at least one representative of the order (or family) of each of the two species. This way, if we observe a correlation, it is not because the respective species exist in a given order (or family). A purely practical reason for applying such a constraint is that many data sets are compiled from literature that is typically organised taxonomically, potentially creating a pseudo-absence effect. Therefore, the dogs and horses might be known from a site, but not the deer, even though they were in fact present, but not relevant for the study that produced the data. By using taxonomic criteria, we have evidence for each included location that some representatives of the order or family of each of the two species is present. Therefore, it is more unlikely that there is pseudo-absence or irrelevant absence effects in that location.

In the example analysis below, where we analyze fossil find sites of large land mammals, we have in fact already implicitly used a kind of taxonomic restriction. Imagine that in our full data set we had find sites having large and small land mammals, respectively. Because we only want to study the effects within large land mammals, we select into the base set only find sites in which large land mammals occur. This way we can rule out the possibility that any correlation we might observe would be due to the differences of large and small land mammals.