extensions - Allow for analysis with priors informed by national analysis of demographic disturbance relationships

[sebdalgarno]

The contractor will extend the bboutools survival and recruitment models to allow for analysis with Bayesian priors informed by national analysis of demographic disturbance relationships.

Josie to provide priors -
The prior means will vary with anthropogenic disturbance (both survival and recruitment) and fire disturbance (only recruitment), so there will need to be some way for users to specify their (static) level of anthropogenic and fire disturbance. That could be done with a single name if you don’t want to add extra arguments e.g. Johnson2020A15F5 could be a row label in a table of priors derived from Science to inform policy: Linking population dynamics to habitat for a threatened species in Canada - Johnson - 2020 - Journal of Applied Ecology - Wiley Online Library that corresponds to 15% anthropogenic disturbance and 5% fire disturbance.

We can have an initial table derived from Johnson et al 2020 and associated explanation ready for you when the contract starts. And we will have another ‘prior set’ to consider later in the year derived from more recent analysis to inform recovery strategy update.

Seb:
need to figure out best way to provide option to user for setting prior -
either 1 single arg that references a table ("johnson2020A15F5" - Johnson 2020, anthropogenic disturbance 15%, fire disturbance 5%)
OR two args (anthropogenic_disturbance , fire_disturbance)
questions: are any combination of values possible? if not then probably makes sense to have one argument)

[sebdalgarno]

updates from Josie:

there is v1 of a function (getBBNationalInformativePriors to get bboutools informative priors implemented in this dev branch of caribouMetrics R package: LandSciTech/caribouMetrics at BbouIntegration

current vignette describing functionality: Comparing the caribouMetrics Beta model to bboutools models • caribouMetrics

Josie needs to make a table of prior parameter values for integer values of anthropogenic and fire disturbance. table probably should live in caribouMetrics to facilitate updates/maintenance.

[sebdalgarno]

will need to access prior values from caribouMetrics. there are several options for dependency:

1. put in Suggests and only require if prior table accessed (i.e. if not already installed)
2. put in Imports and accept expanded dependency chain
3. pull in data in data-raw at package build - but risk going out of sync
4. table of priors lives in bboutools - but need to provide PR merges for updates in timely fashion
5. table of priors lives in new package (caribouData?) with fewer dependencies - but more work for client
6. table of priors lives in new poisson package (bboupriors?) and grant Josie write access to repo

how important is it to avoid this dependency chain?

Imports:
dplyr,
data.table,
sf,
tidyr,
stringr,
purrr,
methods,
matrixStats,
simstudy,
rlang,
terra,
R2jags,
rjags,
survival,
bboutools,
mcmcr,
fitdistrplus

Client prefers that it lives in caribouMetrics.
Likely 1) is the best option unless the feature is very regularly used (including in Shiny app)