Page 1 of 1

Robust Design with Varying Capture Methods

PostPosted: Thu Nov 04, 2021 3:54 pm
by tspaulding
Hello,
I am a novice with Mark Recapture and have what I believe to be a somewhat complicated study design due to permits and client reqs.

I am sampling a waterbody to collect a species of fish and estimate the population size, the waterbody is closed off from other fish, but fish in the waterbody can be removed (death/entrainment) and the fish are breeding within the waterbody. Each year of sampling we are capturing and marking over two days (first day capture, second day recapture) with the understanding that both days will be a recapture opportunity for any fish marked in previous years. My understanding is that this indicates that I should use a Robust Design, with a each of the 8 basic Huggins models for the capture event.

Every year, however, we have to switch off methods of capture, year 1 is electrofishing, year 2 is minnow trapping, year 3 electrofishing, etc. I believe that this therefore has a requirement of accounting for differing "effort" in each sampling year.

For this first year I used a Huggins and Huggins Full Heterogeneity model in RMark to estimate the population size based on the first 6 basic Otis models using the following code (I did not incorporate the full Mtbh model):

Code: Select all
  #Format as encounter history
  data <- data_tmp %>%
    unite(col = ch, -c(Waterbody:PIT_ID), sep = "")%>%
    select(-Species)%>%
    select(PIT_ID, ch, Waterbody)


  dp = process.data(data, model = "Huggins")
  ddl = make.design.data(dp)

  do_analysis = function(data){
    # Defining parameter models
    pdotshared=list(formula=~1,share=TRUE)
    ptimeshared=list(formula=~time,share=TRUE)
    ptime.c=list(formula=~time+c,share=TRUE)
    pdotmixture=list(formula=~mixture,share=TRUE)
    pmixture=list(formula=~mixture)
    cmixture=list(formula=~mixture)
    ptimemixtureshared=list(formula=~time+mixture,share=TRUE)

    # Capture Closed models
    M0=mark(data,model="Huggins", model.parameters=list(p=pdotshared))
    Mb=mark(data,model="Huggins")
    Mt=mark(data,model="Huggins", model.parameters=list(p=ptimeshared))

    # Heterogeneity models
    Mh=mark(data,model="FullHet",model.parameters=list(p=pdotmixture))
    Mbh=mark(data,model="FullHet",model.parameters=list(p=pmixture,c=cmixture))
    Mth=mark(data,model="FullHet",model.parameters=list(p=ptimemixtureshared),adjust=TRUE)

    return(collect.models())
  }

  #results of models and model averaging
  results=do_analysis(data)


Any assistance with the proper identification of appropriate model for next year, help with proper model specification, and any comments on the above script would be helpful!