SPI-IPM Code Manual
About this manual
How to cite
1
Preparing SPI-Birds data for Bayesian analysis
1.1
Nest count data
1.2
Clutch size data
1.2.1
Nest level
1.2.2
Population level
1.3
Fledgling count data
1.3.1
Nest level
1.3.2
Population level
1.4
Mark-recapture data
1.4.1
Individual capture histories
1.4.2
M-array
1.5
Immigrant count data
1.6
Auxiliary data on the sampling process
1.6.1
Nest survey sampling effort
1.6.2
Capture probability proxies
1.7
Organisation for analysis with NIMBLE
2
IPM Construction
2.1
Open population model with 2 age classes
2.1.1
Model description
2.1.2
Code implementation including demographic stochasticity
2.2
Data likelihoods
2.2.1
Nest count data likelihood
2.2.2
Clutch size data likelihoods
2.2.3
Fledgling count data likelihoods
2.2.4
Mark-recapture data likelihood
2.2.5
Immigrant count data likelihood
2.3
Priors and constraints
3
Modelling temporal variation
3.1
Random year variation
3.2
Temporal covariates
3.2.1
Continuous variables
3.2.2
Categorical variables
3.2.3
Imputation of missing covariate values
3.3
Notes on covariate selection
4
IPM Implementation
4.1
Efficient implementation using NIMBLE
4.1.1
Alternative specification of distributions
4.1.2
Vectorized calculations
4.1.3
Custom distributions
4.1.4
Calling NIMBLE from R
4.2
Simulation of initial values
4.3
Test runs and full runs: chains, iterations, burn-in, and thinning
4.4
Trouble-shooting implementation issues
5
Model Assessment
5.1
Assessing chain convergence
5.2
Plotting data vs. predictions
5.3
Comparing estimates from integrated vs. independent analyses
5.4
“Reality check” using stochastic simulations
5.5
Other approaches
6
Visualizing and interpreting direct IPM outputs
6.1
Population trajectories
6.2
Within-population variation in vital rates
6.2.1
Age-class-specific averages
6.2.2
Year-by-year variation
6.3
Between-population variation in vital rates
6.3.1
Population-specific averages
6.3.2
Year-by-year variation
6.4
Covariate effects
7
Follow-up Analyses
7.1
Testing for time-trends
7.2
Testing for density-dependence
7.3
Investigating cross-population covariation
7.4
Quantifying demographic contributions to short term population dynamics
7.4.1
Year-by-year variation in population growth rate (random design LTRE)
7.4.2
Year-to-year differences in population growth rate (fixed design LTRE)
7.5
Quantifying demographic contributions to long-term population trends
7.5.1
Differences in population trajectories between time periods (period design LTRE)
8
Useful extensions and outlook
8.1
Adapting the population model for your species/population
8.1.1
Accounting for multiple broods per bird per year
8.1.2
Altering age structure
8.1.3
Individual heterogeneity beyond age: sex, traits, and more
8.2
Including additional data and informative priors
8.2.1
Including partially observed age information
8.2.2
Making the most of auxiliary knowledge about immigrants/dispersers
8.2.3
Letting published values help with estimation when data is sparse
8.3
Building on the multi-population perspective
8.3.1
Joint analysis of data from several populations
8.3.2
Modelling cross-population covariation
8.3.3
Estimating hyper-parameters in large-scale analyses
8.3.4
Unlocking the secrets of dispersal
References
Published with bookdown
SPI-IPM Code Manual
Chapter 5
Model Assessment
5.1
Assessing chain convergence
5.2
Plotting data vs. predictions
5.3
Comparing estimates from integrated vs. independent analyses
5.4
“Reality check” using stochastic simulations
5.5
Other approaches
Running for additional years and comparing to non-included data, PPCs, etc.