estimates of fidelity for HY age classes

[splash1199]

Hi RMark Users,

I need some clarification on estimates of fidelity in live-dead models. In RMARK: A gentle introduction, fidelity is defined as the probability that a bird that has survived from year i to i+1 has returned to capture site.

Given this definition, why does RMARK estimate fidelity for the hatch-year age class, given that no birds return at t+1 as hatch years?

In my case, I am modelling survival of waterfowl banded as HY, SY, ASY, and I don't think it makes sense to have HY fidelity estimates. But I can't seem to make the model run.

I have provided my code below. I would like to model F as
"~ -1 + sy + asy:sex" but the model will not run without hy included, "~ -1 + hy + sy + asy:sex"

any ideas and or clarification?

Code: Select all

pre.df <- convert.inp("./data/SNGO_BLGO_livedead_1998-2007.inp", group.df = data.frame(age_sex=c("HY Male", "HY Female", "SY Male", "SY Female", "ASY Male", "ASY Female")) , use.comments = TRUE)

#split the age and sex information into two separate columns
pre.df <- tidyr::separate(pre.df, age_sex, c("age", "sex"), " ")
#coerce age and sex into factors
pre.df$age <- as.factor(pre.df$age)
pre.df$sex <- as.factor(pre.df$sex)
```

#process the data by defining the model type as "Burnham" and the groups in the data
pre.proc = process.data(data = pre.df,
   model       = "Burnham",
   groups      = c("age", "sex"),
   age.var     = 1,
   initial.age = c(2, 0, 1))
#ASY (2), HY (0), SY (1)

# make the design data from the process data above
pre.ddl = make.design.data(pre.proc)

#add the correct binning to the design data so that individuals that were marked as HY are SY
#in their second year of life, and ASY in their third year, etc
pre.ddl=add.design.data(data = pre.proc,
   ddl = pre.ddl,
   parameter="S",
   type = "age",
   bins = c(0, 0.5, 1.5, 30),
   right = FALSE,
   name = "age",
   replace = TRUE)
levels(pre.ddl$S$age) = c("HY", "SY", "ASY")
#add a dummy variables; all three age classes
pre.ddl$S$hy     = ifelse(pre.ddl$S$age == "HY", 1, 0)
pre.ddl$S$sy     = ifelse(pre.ddl$S$age == "SY", 1, 0)
pre.ddl$S$asy    = ifelse(pre.ddl$S$age == "ASY", 1, 0)
#add a dummy variables for HY and AHY
pre.ddl$S$adult  = ifelse(pre.ddl$S$age != "HY", 1, 0)
pre.ddl$S$young  = ifelse(pre.ddl$S$age == "HY", 1, 0)

#pre.ddl$S$adultclass = ifelse(pre.ddl$S$age == "HY", "HY", "AHY")


#repeat for p
pre.ddl=add.design.data(data = pre.proc,
   ddl = pre.ddl,
   parameter="p",
   type = "age",
   bins = c(0, 0.5, 1.5, 30),
   right = FALSE,
   name = "age",
   replace = TRUE)
levels(pre.ddl$p$age) = c("HY", "SY", "ASY")
#add a dummy variables; all three age classes
pre.ddl$p$hy     = ifelse(pre.ddl$p$age == "HY", 1, 0)
pre.ddl$p$sy     = ifelse(pre.ddl$p$age == "SY", 1, 0)
pre.ddl$p$asy    = ifelse(pre.ddl$p$age == "ASY", 1, 0)
#add a dummy variables for HY and AHY
pre.ddl$p$adult  = ifelse(pre.ddl$p$age != "HY", 1, 0)
pre.ddl$p$young  = ifelse(pre.ddl$p$age == "HY", 1, 0)

#repeat for F
pre.ddl=add.design.data(data = pre.proc,
   ddl = pre.ddl,
   parameter="F",
   type = "age",
   bins = c(0, 0.5, 1.5, 30),
   right = FALSE,
   name = "age",
   replace = TRUE)
levels(pre.ddl$F$age) = c("HY", "SY", "ASY")
#dummy variables for all three age classes
pre.ddl$F$hy     = ifelse(pre.ddl$F$age == "HY", 1, 0)
pre.ddl$F$sy     = ifelse(pre.ddl$F$age == "SY", 1, 0)
pre.ddl$F$asy    = ifelse(pre.ddl$F$age == "ASY", 1, 0)
#dummy variables for HY and AHY
pre.ddl$F$adult  = ifelse(pre.ddl$F$age != "HY", 1, 0)
pre.ddl$F$young  = ifelse(pre.ddl$F$age == "HY", 1, 0)

#repeat for r
pre.ddl=add.design.data(data = pre.proc,
   ddl = pre.ddl,
   parameter="r",
   type = "age",
   bins = c(0, 0.5, 30),
   right = FALSE,
   name = "age",
   replace = TRUE)
levels(pre.ddl$r$age) = c("HY", "AHY")
#dummy variables for HY and AHY
pre.ddl$r$adult  = ifelse(pre.ddl$r$age != "HY", 1, 0)
pre.ddl$r$young  = ifelse(pre.ddl$r$age == "HY", 1, 0)

test2 = RMark::mark(data = pre.proc,
   ddl = pre.ddl,
  model.parameters = list(
    S = list(formula = ~ -1 + age + time),
    p = list(formula = ~ -1 + sy + asy:time),
    F = list(formula = ~ -1 + hy + sy + asy:sex),
    r = list(formula = ~ -1 + young + adult:sex)),
   invisible = FALSE,
   model = "Burnham")
```

[cooch]

Hi RMark Users,

I need some clarification on estimates of fidelity in live-dead models. In RMARK: A gentle introduction, fidelity is defined as the probability that a bird that has survived from year i to i+1 has returned to capture site.

Given this definition, why does RMARK estimate fidelity for the hatch-year age class, given that no birds return at t+1 as hatch years?

I'll let Jeff sort through the RMark bits, but I suspect you're misunderstanding the fidelity parameter (F). The complement, 1-F, is the probability of *permanent* emigration, not temporary (this is noted in several places in chapter 9). So, flipping the logic, if a marked individual is in the sample AHY, then it was also in the sample HY. You may not live encounter HY individuals, but in theory, a dead recovery can occur anywhere, and so, F for HY can be estimated. [If a dead recovery can't occur anywhere, but (say) occurs only in the areas where you also live mark, then the definition of fidelity means something else.]

Again, (1-F) means *permanent* emigration. It seems as if you're thinking temporary emigration (i.e., that a HY individual leaves the live encountere sampling area, and then returns as AHY). Fine, if thats the case, the p for live encounter for HY is fixed to 0, and you move on from there.