The objectives of this work are: (1) to define spider guilds for all extant families worldwide; (2) test if guilds defined at family level are good surrogates of species guilds; (3) compare the taxonomic and guild composition of spider assemblages from different parts of the world; (4) compare the taxonomic and functional diversity of spider assemblages and; (5) relate functional diversity with habitat structure. Data on foraging strategy, prey range, vertical stratification and circadian activity was collected for 108 families. Spider guilds were defined by hierarchical clustering. We searched for inconsistencies between family guild placement and the known guild of each species. Richness and abundance per guild before and after correcting guild placement were compared, as were the proportions of each guild and family between all possible pairs of sites. Functional diversity per site was calculated based on hierarchical clustering. Eight guilds were discriminated: (1) sensing, (2) sheet, (3) space, and (4) orb web weavers; (5) specialists; (6) ambush, (7) ground, and (8) other hunters. Sixteen percent of the species richness corresponding to 11% of all captured individuals was incorrectly attributed to a guild by family surrogacy; however, the correlation of uncorrected vs. corrected guilds was invariably high. The correlation of guild richness or abundances was generally higher than the correlation of family richness or abundances. Functional diversity was not always higher in the tropics than in temperate regions. Families may potentially serve as ecological surrogates for species. Different families may present similar roles in the ecosystems, with replacement of some taxa by other within the same guild. Spiders in tropical regions seem to have higher redundancy of functional roles and/or finer resource partitioning than in temperate regions. Although species and family diversity were higher in the tropics, functional diversity seems to be also influenced by altitude and habitat structure.

Within behavioural research, non‐normally distributed data with a complicated structure are common. For instance, data can represent repeated observations of quantities on the same individual. The regression analysis of such data is complicated both by the interdependency of the observations (response variables) and by their non‐normal distribution. Over the last decade, such data have been more and more frequently analysed using generalized mixed‐effect models. Some researchers invoke the heavy machinery of mixed‐effect modelling to obtain the desired population‐level (marginal) inference, which can be achieved by using simpler tools—namely by marginal models. This paper highlights marginal modelling (using generalized estimating equations [GEE]) as an alternative method. In various situations,GEEcan be based on fewer assumptions and directly generate estimates (population‐level parameters) which are of immediate interest to the behavioural researcher (such as population means). Using four examples from behavioural research, we demonstrate the use, advantages, and limits of theGEEapproach as implemented within the functions of the ‘geepack’ package in R.