I decided on the Robust Design model. I have four primary periods (2018 – 2021), and three secondary periods in each primary period (April-May, June-July, Aug-Sept). In 2020, I could not conduct surveys for the first secondary period (April-May). So, I created a dummy occasion to make sure the time intervals are equal. I also fixed the p and c for that occasion at 0.
- Code: Select all
#Set time intervals between primary and secondary occasions
time.intervals = c(0,0,1,0,0,1,0,0,1,0,0)
#Create the processed dataframe and design data
rd = process.data(data = rd.txt,
model = "RDHuggins",
groups = "sex",
time.intervals = time.intervals)
ddl = make.design.data(rd)
#Fix the capture and recapture probability for 2020 THIRD PRIMARY AND FIRST SECONDARY occasion to zero (it is a dummy occasion).
ddl$p$fix=NA
ddl$p$fix[which(ddl$p$time==1 & ddl$p$session == 3)] = 0
ddl$c$fix=NA
ddl$c$fix[which(ddl$c$time==1 & ddl$c$session == 3)] = 0
In terms of survival estimates, I considered several models (constant/transience/sex/additive and interactive models). One thing to note is that in 2020, the population experienced a mass mortality event, and I found many unmarked dead female individuals. (Because they are unmarked I cannot add them to my analysis.)
- Code: Select all
#Include the mass mortality year
ddl$S$mme = 0
ddl$S$mme[ddl$S$time == 2] = 1
I also found no evidence of trap dependence based on U-CARE. So my p and c estimates were shared.
- Code: Select all
p.effort = list(formula = ~effort, share = TRUE)
p.season = list(formula = ~season, share = TRUE)
p.constant = list(formula = ~1, share = TRUE)
I also considered markovian, random, and no emigration.
When I ran different models, the best model, S(~1)Gamma''(~1)Gamma'(~1)p(~effort)c() and models that were not very well supported, gave me weird abundance estimates for 2020. There was an upward spike for females, and a downward spike for males and unknowns (basically contradicting what actually happened.) However, when I ran the code again without fixing the p and c for 2020 first secondary occasion, the results appear to resemble what occurred in the field. So I think the obtained estimates are likely a result of my recapture probability and the number caught for each group.
M(t+1):
Group 1 = sexFemale
66 56 54 47
Group 2 = sexMale
69 54 26 65
Group 3 = sexUnknown
38 22 18 49
- Code: Select all
Real Function Parameters of { S(~1)Gamma''(~1)Gamma'(~1)p(~effort)c() }
95% Confidence Interval
Parameter Estimate Standard Error Lower Upper
-------------------------- -------------- -------------- -------------- --------------
1:S gFemale c1 a0 t1 0.6003368 0.0570538 0.4852039 0.7053569
2:p gFemale s1 t1 0.1102262 0.0137007 0.0860983 0.1400791
3:p gFemale s3 t2 0.0829668 0.0138253 0.0595882 0.1144017
4:Gamma'' gFemale c1 a 0.0000000 0.0000000 0.0000000 0.0000000 Fixed
- Code: Select all
Estimates of Derived Parameters
Population Estimates of { S(~1)Gamma''(~1)Gamma'(~1)p(~effort)c() }
95% Confidence Interval
Grp. Sess. N-hat Standard Error Lower Upper
---- ----- -------------- -------------- -------------- --------------
1 1 223.29864 33.712874 169.83246 304.29601
1 2 189.80800 29.862809 142.85885 262.13419
1 3 340.29669 68.999878 233.70275 510.11873
1 4 205.85249 41.044656 143.51581 308.45058
2 1 233.87057 34.988066 178.26660 317.77048
2 2 183.02914 29.066215 137.42636 253.55946
2 3 163.84655 39.422568 105.56571 264.81735
2 4 284.68961 53.419421 202.33903 416.41887
3 1 128.79828 22.573274 94.185223 184.73482
3 2 74.567427 15.678404 51.663858 115.15492
3 3 113.43223 30.494711 69.793866 193.83762
3 4 214.61217 42.429045 150.03482 320.46475
Any advice is appreciated.