To construct a database for processing in capture-recapture frame
models, we organized the life table obtained from observations collected
during the experiment. We considered recapture events every 5 days, with
a reference to the duplication time of 7 days for a population under
optimal conditions (Gotelli, 1995). This resulted in a life history
database of 2197 samples with 35 capture-recapture occasions of same-age
individuals. As our experimental population is closed, we applied a
single state Cormack-Jolly-Seber (CJS) model, followed by a multi-state
model accounting for different developmental stages (Larvae, Pupa, and
Adult: L, P, and A respectively). Detection probability remained
constant throughout the experiment under controlled conditions. The CJS
models applied were: a) CJS model (ϕ, p) with constant detection and
survival; b) Time-varying survival and constant detection (ϕt, p); and,
c) Constant detection and survival dependent on treatment (ϕtreat, p).
In all cases, model selection was based on lower WAIC. In multi-state
models, we considered developmental states (L, P, and A) plus dead as
states (four states in total), with transitions only possible from L to
P and from P to A, and from any state to dead. Detection was considered
constant, and we applied two different models to estimate survival: a) A
multistate model with constant detection (ϕL, ϕP, ϕA, ψLP, ψPA, pL, pP,
pA); and, b) A multistate model with constant detection and
treatment-dependent survival (ϕL, ϕP, ϕA, ψLP, ψPA, pL, pP, pA, βtreat).
The latter aimed to identify a differential effect of treatments on the
different developmental phases.
Results are indicated as mean ± STD unless indicated otherwise.