Statistical analyses
We performed all statistical analyses using R version 3.4.2 (R Core Team
2016) and created all figures using the ‘ggplot2’ package with R
(Wickham 2016). We visually inspected all diagnostic plots (qqplots and
plots of the distribution of the residuals against fitted values) to
validate model normality. In cases where model assumptions of normality
were not, we assessed the normality of the response variable using a
Shapiro-Wilk test and transformed variables as needed until normality of
models was met. Only the best fitting models are reported here.
To assess differences in the viscosity of female reproductive fluids
between polyandrous and monogamous species, we used our dataset of
individually tracked particles and pooled these data across species with
shared mating systems (categorized as either monogamous or polyandrous).
We excluded data that were ±2SD of the mean within each sample because
they represented clear outliers. All viscosity values were
log-transformed to improve model assumptions of normality. We assessed
differences in the viscosity of female reproductive fluids in the uterus
and in the oviduct using separate linear models (LM), with log viscosity
as the response variable and mating system as the predictor variable. We
also analyzed fluidic differences in each tract region within each
species using separate linear models, with log viscosity as the response
variable and the region of the female reproductive tract as the
predictor variable. Last, we used the data sets from three focal samples
(P. eremicus , P. gossypinus , and P. polionotus ) to
statistically compare viscosity measurements to their dyed mineral oil
controls using separate linear models, with log viscosity as the
response variable and the sample type as the predictor variable.
Post-hoc pairwise comparisons were made using Tukey HSD adjustments for
multiple comparisons using the ‘LSmeans’ R package (Lenth 2016).
To assess differences in the pH and calcium levels of female
reproductive fluids between polyandrous and monogamous species, we used
separate linear models for each region of the tract, with either pH or
the log calcium measurements included as the response variable and
mating system included as the explanatory variable. Last, we used paired
t-tests to conduct pairwise comparisons for pH and calcium measurements
from each region of the tract within each species to control for
differences among individual females.