Appendix C. Model selection results for the complete model set fit to data from partially migrating red-spotted newts, Notophthalmus viridescens.
Model selection was based on quasi-likelihood Akaike Information Criteria (QAICc), where the model with the lowest QAIC value was considered ‘best’. The overdispersion parameter was estimated using median ĉ methods implemented in program MARK (ĉ = 3.46). Additional model selection statistics for each model include relative QAICc (ΔQAICc), number of parameters (K), Akaike weight (w), and -2log likelihood (-2Log(L)).
TABLE C1. Model selection results for eight different capture probability structures representing all possible combinations for time × sex × pond using the global structure for survival and transition probabilities, φ(time × sex × pond × tactic) ψ(time × sex × pond × tactic). Time variation in capture probability was modeled either with annual variation among spring samples (t2,4,6=8) or with spring 2006 being different from subsequent spring seasons (t2,4=6=8). Capture probability was always modeled separately for migrants and residents (tactic) due to the different sampling methods used (drift fence, seine, or aquatic traps). Recall for all relevant states (Appendix B).
|p(t2,4=6=8 × sex × tactic)||7606.44||0||116||0.59||25492.15|
|p(t2,4=6=8 × tactic)||7608.13||1.69||114||0.25||25512.24|
|p(t2,4,6=8 × sex × tactic)||7610.08||3.64||120||0.10||25476.19|
|p(t2,4,6=8 × tactic)||7610.89||4.45||117||0.06||25500.41|
|p(t2,4=6=8 × pond × tactic)||7617.55||11.12||123||0.00||25480.63|
|p(t2,4=6=8 × sex × pond × tactic)||7618.71||12.27||127||0.00||25456.02|
|p(t2,4,6=8 × pond × tactic)||7623.53||17.09||129||0.00||25458.41|
|p(t2,4,6=8 × sex × pond × tactic)||7626.93||20.50||135||0.00||25427.21|
TABLE C2. Candidate model set with combinations of time × sex × pond × tactic (migratory state) using parallel parameter structures for apparent survival probability, φ, and transition probability, ψ. Time was modeled across years for specific seasons. Yearly variation in φ and ψ was only included for overwinter survival probability () and spring to fall transitions between migrating and residency states (). The same parameters for the opposite seasons were modeled without differences between years (i.e., and ). Variations in apparent survival and transition probabilities among migratory state were modeled with : (1) parameter differences between migrants, residents, and newly captured individuals (tactic), (2) parameters for migrants and newly captured individuals in the spring set equal and parameters for residents and maturing efts in the fall set equal (tacticnew=recap) or (3) no differences between migratory states. Density estimates were added to the top ranked model at t = 2,4,6 as a time-specific pond covariate to test the influence of newt density within the pond on transitions between migratory states.
|t1,3,5,7 × sex × pond × tacticnew=recap||density + sex × tacticnew=recap||7566.88||0||74||0.9998||25653.45|
|1||t1,3,5,7 × sex × pond × tacticnew=recap||t2,4,6 × sex × pond × tacticnew=recap||7584.00||17.12||92||0.0002||25585.42|
|2||sex × tactic||sex × tactic||7599.55||32.67||38||0||26019.14|
|3||sex × tacticnew=recap||sex × tacticnew=recap||7601.70||34.82||32||0||26068.47|
|4||t1,3,5,7 × sex × tacticnew=recap||t2,4,6 × sex × tacticnew=recap||7607.97||41.09||52||0||25950.26|
|5||t1,3,5,7 × sex × pond × tactic||t2,4,6 × sex × pond × tactic||7611.40||44.52||116||0||25509.57|
|6||t1,3,5,7 × sex × tactic||t2,4,6 × sex × tactic||7614.17||47.29||64||0||25887.42|
|7||t1,3,5,7 × sex × pond||t2,4,6 × sex × pond||7656.80||89.92||60||0||26063.00|
|9||t1,3,5,7 × sex||t2,4,6 × sex||7671.09||104.21||36||0||26280.53|
|10||pond × tactic||pond × tactic||7758.29||191.41||38||0||26568.14|
|11||pond × tacticnew=recap||pond × tacticnew=recap||7764.95||198.07||32||0||26633.06|
|12||t1,3,5,7 × tacticnew=recap||t2,4,6 × tacticnew=recap||7805.56||238.68||32||0||26773.51|
|15||t1,3,5,7 × tactic||t2,4,6 × tactic||7811.79||244.91||38||0||26753.17|