Nest Survival

1
Nest Survival

• A special case of known fate analysis

2
Nest Survival Analysis

• Used to model nest data
• Also useful for messy known fate data
• Data that doesnt have all observations at
  specified occasions
• Used to model Daily Survival Rate (DSR)
• Often used to estimate nest survival (success)
• Probability at least one egg in a nest hatches
• Model, when during an interval, mortality
  occurred
• Generally, not used for modeling
  interval-specific survival

3
Why did nest survival analysis evolve?

• Naive estimator was biased
• Apparent survival nests destroyed/ found
• Marked population not representative of unmarked
• More likely to find successful nests
• Positive bias in nest survival, not DSR for the
  period during sampling
• Solution Use nests found at the start of lying
  or incubation
• Not logistically feasible
• Loss of information

4
Mayfield

• Sampling Unit is exposure days, not nests
• DSR (E f)/E
• Where, E exposure days and f failures
• Can this be used for nest survival?
• Maybe
• Inference still needs to be limited to period of
  sampling

5
Bias

• Unbiased estimates of DSR for period of sampling
• How can we use Mayfield to correct problem?

6
Mayfield An Example

• Truth NS 2/10 0.2
• Apparent NS
• Day 3 0.33
• Day 9 0.66
• Mayfield
• mid-point assumption
• DSR (3) 0.888, NS 0.24
• DSR (9) 0.18, NS 0.18
• DSR is not constant

7
Why is Mayfield Imperfect?

• Mid-point assumption
• Constant DSR
• Adhoc
• No theoretical basis or measure of precision
• Improvements
• Johnson et al. and Bart and Robson
• MLE
• 40 assumption, later MLE of time of failure
  during interval
• Variance estimator

8
More Changes

• Other improvements
• Rotella et al. (2000) Observer Effects
• Stanley (2000) Stage-specific survival
• Dinsmore et al. (2002) Software, Covariates,
  Link Functions
• Rotella et al. (2004), Shaffer (2004) More
  general framework, SAS

9
DSR in Generalized Linear Models (GLM)

• Less emphasis on nest survival
• Generally assume that nests are found during all
  stages no Mayfield correction
• Modeling relationship between DSR and covariates
• Data
• Interval between visits, doesnt need to be
  constant
• Fate of nest at end of visits

10
Implementation in MARK and Complex Models in SAS

• Data
• Day nest was found (i)
• Last day checked alive (j)
• Last day checked (k)
• Fate (f 0successful, 1failed)
• Number of nests with that CH

• Thus
• jltk
• If jk successful
• klthatch date
• jgti

11
Input
/Day found, Day before last, Last day, Fate
(0survived, 1depredated),number, occasions6
groups2 /Nest survival group11 2 2 0 811
1 2 1 21 3 3 0 301 1 3 1 11 4 4 0 611 5 5
0 371 1 5 1 11 6 6 0 111 1 6 1 11 7 7 0
211 1 7 1 2
/ Bart and Robson Nest Survival Data, in form of
encounter histories /10100000000000 81
1211000000000000 2 010101000000000 30
710010000000000 1 110101010000000 61
12 10000100000000 0 310101010100000 37
1710000001000000 1 110101010101000 11
2410000000010000 1 210101010101010 21
3110000000000100 2 2
12
Design

• Tradeoffs between the number of nests/monitoring
  frequency
• More individuals should provide more precision
  than more frequent visits
• If the study spans several life stages, then
  monitoring should be frequently enough to
  estimate when the transition between stages
  occurred
• Record nest age and date during a visit because
  these covariate is often explanatory and may lead
  to higher precision
• Therefore, frequency of visits may be variable
  depending on development stage
• Disturbance at the nest site may reduce survival
• Increasing length of time between visits
• Varying length of time between visits may allow
  estimation of how much effect observers have on
  survival
• Temporal and spatially stratified sampling
  schemes may be useful for maximizing the
  inference scope for these studies by increasing
  the probability of detecting nests that fail
  early and therefore reducing heterogeneity in the
  sample

13
Models

• Numerous
• Similar in structure to known fate with
  structured occasions
• More verbose to write in MARK

14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)