Effect-size and study trait variable extraction
We recorded the following information from each study to allow direct
comparison of effect sizes, test the effect of study features
(moderators) on this effect, and control for variation between studies:
host/prey taxa to test for a phylogenetic trend in our models; parasite
type (macroparasite, microparasite, or parasitoid), study design
(observational or experimental), predator interaction type (all or
non-consumptive), and predator-spreader identity (predator-spreader or
not) for inclusion in mixed effects models (MEMs) testing the effect of
these moderators on effect sizes. The majority of studies (45 of 50)
were composed of a binary comparison of a parasite response across two
levels of predation. Most studies were analyzed using multivariate
statistics which makes statistical comparison of effect sizes across
studies challenging
(Borenstein et
al. 2017). For this reason, we extracted the mean parasite response
value, sample size, and measure of variation (typically SE, SD, or 95%
CI) from the text or figures of each of these studies and calculated the
standardized mean difference (Hedges g) using the escalc function
in the R package metafor
(Viechtbauer 2010). A
small minority of studies (5 of 50) reported parasite responses over a
range of predation pressures. We converted responses from 3 of these
studies to binary effect sizes by using raw data provided to compare the
mean parasite response for samples in the first quartile of predator
abundance to those in the 4th quartile of predator abundance. We
excluded studies from further analysis if sufficient data for this
procedure were not provided. Following this protocol we extracted 193
effect sizes from 48 studies (Table 1).
Not all effect sizes contain the same type of information because of
differences in the biology of parasites and in the associated response
metric. For our study, we grouped effect sizes into 2 broad categories
based on the parasite response that was measured: (i) the number or
proportion of hosts infected (quantified as prevalence, number or
density of infected individuals, or disease induced mortality rate;n = 89 effect sizes from 22 different studies, Table 2) and (ii)
the number of parasites in an average individual (quantified as parasite
intensity or parasite load; n = 61 effect sizes from 19 different
studies). Because we expected that predators would have different
effects on prevalence and intensity measures (for example a small amount
of selective predation on a population with highly aggregated parasites
may have a large effect on mean intensity but a small effect on
prevalence), we analyzed these responses separately. Another distinction
we made was to separate parasites from parasitoids. Parasitoids behave
like both predators and parasites over the course of their life-cycle.
Adult parasitoids are free-living flies and wasps that lay eggs on live
hosts, but the juvenile parasitoids that hatch from these eggs are
obligately parasitic and typically lethal to the host. Consequently, the
effect of predators on parasitoids in prey may result from different
processes than the effects on typical parasites. For this reason, we
analyzed parasitoids (n = 43 effect sizes from 11 different
studies) separately from parasites.