Hi again:
The best model for the trapping data includes a behavioral response and a time trend on detectability. Initially when I was beginning to analyze the data I ran models that included individual variables, like sex and age class, but it seems that there is no heterogeneity in catchability. Faecal DNA, on the other hand, is susceptible to other variables, that are included in the best model for the noninvasive genetics (genotyping success and the area of each searched habitat patch). So, I think that the heterogeneity in catchability is possibly not the reason for the lower density estimates obtained with live-trapping. I´ll post the print of these models in the end so you can have a look if you want.
The difference in AICc was high (12) between HEX and EX. I’ll post the print of these models also. I must discuss with my advisers the need to re-run the models using HEX instead of EX. There are more than 100 models that need to be re-run, if this is the case, and at least one month of computation time, with several models running parallel (some of these models took almost a week to run). At this point, when the paper is almost ready to submit, we need to evaluate if these extra time and effort is worthwhile.
Once again, thank you so much for your help.
Helena
Best model for the live-trapping data- Code: Select all
> print(Model58b.V3, deriv = TRUE)
secr.fit(capthist = CaptHistA.V3, model = list(D ~ 1, g0 ~ T +
B, sigma ~ B), mask = MaskA.V3, CL = FALSE, detectfn = 2,
start = c(2.8063528, -5.6620229, 0.1937767, 2.7593377, 1.8213741,
-0.1933307), method = "Nelder-Mead", verify = TRUE, trace = TRUE)
secr 2.8.2, 09:57:21 01 Jul 2014
Detector type single
Detector number 371
Average spacing 4.794238 m
x-range 521534.6 522328 m
y-range 4178202 4179022 m
Usage range by occasion
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
min 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
max 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
N animals : 31
N detections : 81
N occasions : 20
Mask area : 11.1104 ha
Model : D~1 g0~T + B sigma~B
Fixed (real) : none
Detection fn : exponential
Distribution : poisson
N parameters : 6
Log likelihood : -372.8012
AIC : 757.6023
AICc : 761.1023
Beta parameters (coefficients)
beta SE.beta lcl ucl
D 2.8083653 0.189386184 2.4371752 3.1795554
g0 -5.6287007 0.015227904 -5.6585468 -5.5988545
g0.T 0.1916259 0.006550623 0.1787869 0.2044649
g0.BTRUE 2.7079777 0.017825394 2.6730406 2.7429149
sigma 1.8426776 0.010155018 1.8227741 1.8625810
sigma.BTRUE -0.2197838 0.013723621 -0.2466816 -0.1928860
Variance-covariance matrix of beta parameters
D g0 g0.T g0.BTRUE sigma sigma.BTRUE
D 0.0358671266 -1.394867e-04 -3.642875e-04 1.416620e-04 -1.035854e-04 1.092309e-04
g0 -0.0001394867 2.318891e-04 -1.002422e-06 -2.318122e-04 -1.625975e-07 3.123691e-07
g0.T -0.0003642875 -1.002422e-06 4.291066e-05 7.311885e-07 -3.942257e-06 3.252029e-06
g0.BTRUE 0.0001416620 -2.318122e-04 7.311885e-07 3.177447e-04 2.490738e-07 -5.308419e-07
sigma -0.0001035854 -1.625975e-07 -3.942257e-06 2.490738e-07 1.031244e-04 -1.029646e-04
sigma.BTRUE 0.0001092309 3.123691e-07 3.252029e-06 -5.308419e-07 -1.029646e-04 1.883378e-04
Fitted (real) parameters evaluated at base levels of covariates
link estimate SE.estimate lcl ucl
D log 16.582788101 3.168923e+00 11.44067739 24.036064629
g0 logit 0.003580376 5.432641e-05 0.00347546 0.003688447
sigma log 6.313420226 6.411455e-02 6.18900352 6.440338060
Best model for noninvasive genetics- Code: Select all
> print(ModelR52.V4b, deriv = TRUE)
secr.fit(capthist = CaptXY1.V4, model = list(D ~ 1, g0 ~ sa +
log(tarea_m2), sigma ~ 1), mask = Mask1.V4, CL = FALSE, detectfn = 2,
start = c(3.22372542, 4.64804125, 0.04209972, -0.95883076,
1.63094581), method = "Nelder-Mead", verify = TRUE, trace = TRUE)
secr 2.8.2, 19:01:18 20 Dez 2014
Detector type polygon
Number vertices 265
Number polygons 21
Total area 9.867063 ha
x-range 521529.8 522347.7 m
y-range 4178147 4179094 m
N animals : 65
N detections : 115
N occasions : 4
Count model : Poisson
Mask area : 11.1104 ha
Model : D~1 g0~sa + log(tarea_m2) sigma~1
Fixed (real) : none
Detection fn : exponential
Distribution : poisson
N parameters : 5
Log likelihood : -204.8343
AIC : 419.6686
AICc : 420.6856
Beta parameters (coefficients)
beta SE.beta lcl ucl
D 3.22638369 0.155871514 2.92088114 3.53188624
g0 4.65592872 1.501422374 1.71319494 7.59866250
g0.sa 0.04207808 0.006365284 0.02960236 0.05455381
g0.log(tarea_m2) -0.95982951 0.182463477 -1.31745135 -0.60220767
sigma 1.63177974 0.114332966 1.40769125 1.85586824
Variance-covariance matrix of beta parameters
D g0 g0.sa g0.log(tarea_m2) sigma
D 0.0242959287 0.039416865 2.658278e-04 -7.811774e-03 -1.743334e-04
g0 0.0394168650 2.254269144 -2.288354e-03 -2.667201e-01 3.028020e-02
g0.sa 0.0002658278 -0.002288354 4.051684e-05 3.523713e-05 8.706982e-06
g0.log(tarea_m2) -0.0078117741 -0.266720089 3.523713e-05 3.329292e-02 -3.309909e-03
sigma -0.0001743334 0.030280196 8.706982e-06 -3.309909e-03 1.307203e-02
Fitted (real) parameters evaluated at base levels of covariates
link estimate SE.estimate lcl ucl
D log 25.18840301 3.95012304 18.55763209 34.1883945
g0 log 0.07241937 0.02625803 0.03636664 0.1442137
sigma log 5.11296640 0.58649624 4.08650977 6.3972502
Null model for polygon detectors, using EX- Code: Select all
> print(ModelR1.V4, deriv = TRUE)
secr.fit(capthist = CaptXY1.V4, model = list(D ~ 1, g0 ~ 1, sigma ~
1), mask = Mask1.V4, CL = FALSE, detectfn = 2, start = NULL,
method = "BFGS", verify = TRUE, trace = TRUE)
secr 2.10.4, 22:41:08 12 Mar 2017
Detector type polygon
Number vertices 265
Number polygons 21
Total area 9.867063 ha
x-range 521529.8 522347.7 m
y-range 4178147 4179094 m
N animals : 65
N detections : 115
N occasions : 4
Count model : Poisson
Mask area : 11.1104 ha
Model : D~1 g0~1 sigma~1
Fixed (real) : none
Detection fn : exponential
Distribution : poisson
N parameters : 3
Log likelihood : -243.6708
AIC : 493.3416
AICc : 493.7351
Beta parameters (coefficients)
beta SE.beta lcl ucl
D 2.4809079 0.1391105 2.208256 2.7535594
g0 -0.9366488 0.1337347 -1.198764 -0.6745336
sigma 1.5666668 0.1027948 1.365193 1.7681410
Variance-covariance matrix of beta parameters
D g0 sigma
D 0.0193517297 -0.008350741 -0.0004722569
g0 -0.0083507409 0.017884971 0.0027955625
sigma -0.0004722569 0.002795562 0.0105667740
Fitted (real) parameters evaluated at base levels of covariates
link estimate SE.estimate lcl ucl
D log 11.9521105 1.6707404 9.0998353 15.698410
g0 log 0.3919391 0.0526511 0.3015667 0.509394
sigma log 4.7906535 0.4937581 3.9164777 5.859949
Null model for polygon detectors, using HEX- Code: Select all
> print(ModelR1.V4Z, deriv = TRUE)
secr.fit(capthist = CaptXY1.V4, model = list(D ~ 1, g0 ~ 1, sigma ~
1), mask = Mask1.V4, CL = FALSE, detectfn = 16, start = NULL,
method = "BFGS", verify = TRUE, trace = TRUE)
secr 2.10.4, 05:26:57 08 Mar 2017
Detector type polygon
Number vertices 265
Number polygons 21
Total area 9.867063 ha
x-range 521529.8 522347.7 m
y-range 4178147 4179094 m
N animals : 65
N detections : 115
N occasions : 4
Count model : Poisson
Mask area : 11.1104 ha
Model : D~1 lambda0~1 sigma~1
Fixed (real) : none
Detection fn : hazard exponential
Distribution : poisson
N parameters : 3
Log likelihood : -249.7629
AIC : 505.5259
AICc : 505.9193
Beta parameters (coefficients)
beta SE.beta lcl ucl
D 2.531826 0.1418696 2.253766 2.809885
lambda0 -1.086550 0.1299047 -1.341158 -0.831941
sigma 1.526036 0.1170496 1.296623 1.755449
Variance-covariance matrix of beta parameters
D lambda0 sigma
D 0.0201269873 -0.008856176 -0.0002039613
lambda0 -0.0088561763 0.016875236 0.0025374188
sigma -0.0002039613 0.002537419 0.0137006034
Fitted (real) parameters evaluated at base levels of covariates
link estimate SE.estimate lcl ucl
D log 12.5764453 1.79323091 9.5235369 16.6080077
lambda0 log 0.3373786 0.04401262 0.2615426 0.4352037
sigma log 4.5999069 0.54026658 3.6569267 5.7860451