Data Analysis
All data analysis was carried out in R. A principal component analysis was conducted on all harmonic coefficients derived from the Fourier analysis, in order to characterise intraspecific and interspecific variation in pedipalp shape complexity. The resulting PC scores were subject to clustering analysis in order to determine the degree to which species may be correctly identified on the basis of Fourier outlines. PC scores were scaled to have the same mean and standard deviation before being fed into the clustering algorithm. An unsupervised k-means clustering was applied; k-means was instructed to split the dataset into 11 clusters, the same as the number of species, in order to determine if clustering would split down species lines. A pairwise MANOVA was also conducted in ‘Momocs’.
To test the relationship between pedipalp shape complexity and length, species-average values of ‘gross complexity’ were calculated. Due to uncertainty in amblypgyid phylogeny, ordinary least squares regression was then carried out between measures of complexity and pedipalp length for each segment individually. Pedipalp tibia length is considered a standard metric of pedipalp length and is used in multiple studies (Weygoldt, 2000; Prendini, Weygoldt & Wheeler, 2005; McArthur et al., 2018). As such, this was the measure of pedipalp length used herein. A Shapiro-Wilk test confirmed that shape complexity metrics for both segments followed a normal distribution.
Tests for sexual dimorphism in shape complexity were carried out using a nested ANOVA between measures of complexity and sex, with species included as a random effect. Shapiro-Wilk tests confirmed shape complexity in both metrics was normally distributed within species, and species standard deviations were very similar in all species.
A paired Wilcoxon rank-sum test was used to investigate differences in shape complexity between segments, as data were non-normally distributed. Specimens that did not appear in both the tibia and femur dataset were excluded from this analysis, resulting in a total sample size of 75.