tempo_mcmc: tempo_mcmc
In vlandau/tempo: Temporally-Explicit Models for Phenological Observations

MCMC sampler for the discrete-time binary state-transition model.

tempo_mcmc(
  y,
  covariates,
  beta_start,
  n_iter = 1000,
  n_burnin = 1000,
  n_chains = 3,
  n_workers = 1,
  pp_checks = c("mean", "variance"),
  beta_prior_mean = NULL,
  beta_prior_sd = NULL,
  random_effects = NULL,
  group_ids = NULL,
  sd_eps_start = NULL,
  correlated = TRUE,
  monitor_random_effects = FALSE
)

`y`	dataframe; A dataframe with 3 columns: `obs_id` containing a unique ID for each observation; `c1`, the time step of the last observed 0 (or `NA` if a 0 was never observed); and `c2`, the time step of the first observed 1 (or `NA` if a 1 was never observed). In the case of exact observations, e.g. it is known that the even occured on day 42, `c1` should be 41, and `c2` should be 42.
`covariates`	list; A list of `c` covariates matrices, each of dimension `[n,t]`, where n is the number of observations, t is the number of time steps during which state transition could take place, and c is the number of covariates. Use the function `tempo_wrangle` provided in this package to convert long form covariate data to the format necessary for input into `tempo_mcmc`.
`beta_start`	vector; a vector of starting values for model regression coefficients of length `c + 1`, where `c` is the number of covariates. The first value of `beta_start` is for the intercept, and the remaining `c` values correspond to the covariates in the same order as `covariates`.
`n_iter`	integer; The number of mcmc iterations per chain after any burn in. Defaults to 1000.
`n_burnin`	integer; The number of samplers per chain to discard as burn in. Defaults to 1000.
`n_chains`	integer; the number of MCMC chains to sample. Defaults to 3.
`n_workers`	integer; the number of parallel workers to use. Defaults to 1.
`pp_checks`	Character vector; one or both of "mean", "variance". Specifies the statistics for which posterior predictive checks should be run. Set to `NA` to prevent posterior predictive checks from being calculated.
`beta_prior_mean`	vector; Prior means for model regression coefficients in the the same order as beta_start. If `NULL` (the default) then priors means are set to 0.
`beta_prior_sd`	vector; Prior standard deviations for model regression coefficients in the the same order as beta_start. If `NULL` (the default) then priors standard deviations are set to 5. Note that standard deviations that are too restrictive (including the somewhat standard value of 1.5) can influence model fit significantly because posterior parameter magnitudes can be very large.
`random_effects`	vector; a vector of the names of covariates that should be modeled with group-level (random) effects. Names correspond to the names of the list elements of `covariates`. Use "intercept" to specify if the intercept should be modeled with random effects. For example, `random_effects = c("intercept", "precipitation", "temperature")` would specify that random effects should be modeled on the intercept and "precipitation" and "temperature" covariates. Defaults to `NULL` in which case no random effects are modeled.
`group_ids`	vector; If `random_effects != NULL`, indicies specifying to which group each row in `y` belongs. Defaults to `NULL`.
`sd_eps_start`	vector; a vector of starting values for random effect standard deviations. Values correspond to the items in `random_effects`. May be set to `NULL` if `random_effects = NULL`.
`correlated`	Boolean; Should correlations between random effects be modeled? Only applies if there is more than one random effect modeled. Defaults to `TRUE`.
`monitor_random_effects`	Boolean; Defaults to `FALSE`. Return posterior samples for group effects (model parameter `eps`)? Only applies when modeling random effects.