Data analysis
The data was analyzed based on self-regressive linear models with two
parameters of interest, namely the permanence probability of the taxon
in the occurrence of the disturbance \(\left(\psi\right)\), and the
colonization probability in a substrate without previous occurrence of
the taxon \(\left(\theta\right)\). The standard model used has the
following structure:
\begin{equation}
{O_{t}\sim Bern\left(P_{t}\right)\backslash n}{P_{t}=\theta+O_{t-1}\bullet\psi}\nonumber \\
\end{equation}where Ot was the occurrence of the taxon at timet . This model was modified to fit each stage of the experiment.
We used Bayesian inference to adjust the models to the data and to
estimate the credibility intervals (CI; 95%) for the parameters of
interest. All analyses were conducted in R (R Development Core Team
2020), using the runjags package (Denwood 2016). We calculated
the CI using the methodology of re-sampling by relevance, in a Monte
Carlo Markov chain iteration (Gelman et al. 2000, Andrade and Kinas
2008). In all analyses we used 5 parallel chains, burning the first 1000
iterations and sampling every 50 iterations, in a total of 5000 chain
steps. The priori distributions of all analysis were not
informative.
The combinations of the traits of the insects characterize the life
history strategies for the organisms (Poff et al. 2006, Verberk et al.
2013, Sarremejane et al. 2020). We elaborated individual hypotheses for
each life history strategy, because many groups of insects had similar
traits. The hypotheses are related to estimates of the parameters
permanence and colonization and are based on information about how the
functional traits may be affected these parameters (Table 1).