*2.5. Data Analysis*

• Test of phylogenetic signal in services provided by alien woody species

Prior to the analysis, the phylogeny was pruned off the native species. To test whether species used by humans are randomly selected with respect to the services they provide, a matrix of species and service categories was first created for each plant invasion status (alien non-invasive and alien invasive). In this matrix, the 12 categories of services (Table S1) were transformed into binary data, as follows: 1 (if a species provides a given service), and 0 (if not). Then, using the phylogenetic tree of Southern Africa's flora pruned to have only alien species to South Africa, we applied the D statistic [40] on this binary data to assess whether species used for a particular service are phylogenetically more closely related than expected at random (test of phylogenetic signal). D statistic has the advantage of measuring both a phylogenetic signal and its strength. The strength of the signal was interpreted as follows: D < 0 means strong signal; D = 0 means presence of signal under Brownian Motion model; D between 0 and 0.5 means moderate signal; D between 0.5 and 1 means weak signal; D = 1 means no signal; D > 1 means over-dispersion. The statistical significance of the observed D value was tested by comparing the observed D value to 0 (expected value for a phylogenetically conserved pattern under a Brownian Motion model) and 1 (random expectation). The *p* values for significance tests were reported as PBM (giving the result of testing whether D was significantly different from 0) and Prand (giving the result of testing whether D was significantly different from 1). In the scenario of a D value falling between 0 and 1 but being statistically different from 1, this implies that the observed D value shows moderate/weak signal but is non-random. If D value is between 0 and 1 but not statistically different from 1, then the observed value is moderate/weak and not different from random.

• Tests of link between services provided by alien plants and their invasion status

To test if services can be linked to invasion status, we tested whether the diversity of services provided by alien plants (i.e., total number of services for each alien species) correlates with their invasion status. This analysis was carried out by fitting two types of GLM models on "number of services" (response variable) versus "invasion status" (predictor). On one hand, we fitted a Poisson GLM (given the response variable is count data and on the other, we fitted a phyloGLM as implemented in the R library *Phylolm* [41]. The difference between both tests is that the latter corrects for phylogenetic nonindependence of species, allowing us to assess the potential influence of phylogeny on the result reported in the former test.

Finally, we tested whether there was a direct potential link between each service and the invasion status. The test was run by fitting a binomial GLM since invasion status (response variable) was measured as a binary variable (invasive vs. non-invasive following NEMBA).
