tempo_derive: Calculate derived quantities
In vlandau/tempo: Temporally-Explicit Models for Phenological Observations

This function can calulate select derived quantites using the output from tempo_mcmc() for observations chosed by the user. Derived quantities include the pmf and/or cdf of transition time, the transition probabilities for each time step (i.e. the parameter for the tvgeom distribution), the mean predicted transition time, and the standard deviation of the transition time distribution.

tempo_derive(
  draws,
  quantity,
  obs_idx,
  max_samples = 1000,
  parallelize = TRUE,
  n_cores = NULL
)

`draws`	output from `tempo_mcmc()`
`quantity`	character vector; one or more of "probabilities", "pmf", "cdf", "mean", and "sd", specifying which quantities to calculate. Note that if your model included random effects, you must run `tempo_mcmc()` with `monitor_random_effects = TRUE` in order to calculate derived quantities.
`obs_idx`	integer vector; specifies the observations for which derived quantities will be calculated. Indices in `obs_idx` corespond to the rows in the observation data supplied to `tempo_mcmc()`.
`max_samples`	integer; specify the number of samples for which to calculate derived quantites for each chain. Defaults to `1000`. If `max_samples` is less than the number of MCMC samples, each chain in `draws` will be downsampled deterministically to size `max_samples`, resulting in a total of `max_samples * <number of chains>` samples.
`parallelize`	Boolean, should the function be run in parallel. Defaults to `TRUE`.
`n_cores`	Integer, number of cores to use for parallel processing. Defaults to `NULL`. If `parallelize = TRUE`, and `n_corese = NULL` (the dafaults), `tempo_dic()` will use `parallel::detectCores() - 1` as the number of cores.

list of posterior distributions for the derived quantities. "probabilities", "pmf", and "cdf" each return a 3D array with mcmc iterations as rows, time steps as columns, and observations along the 3rd margin. "mean" and "sd" return matrices with mcmc iterations as rows and observations as columns. All objects are returned with named dimensions. The dimension names for observations match the dimention names of the rows for matrices in the list of covariates supplied to tempo_mcmc().

vlandau/tempo documentation built on March 18, 2020, 12:04 a.m.