Functional Dirichlet Process Spatial Temporal Aggregated Predictor in a Linear Model

Functional Dirichlet Process Spatial Temporal Aggregated Predictor in a Linear Model

fdp_staplm(
  formula,
  benvo,
  weights = NULL,
  alpha_a = 1,
  alpha_b = 1,
  sigma_a = 1,
  sigma_b = 1,
  tau_a = 1,
  tau_b = 1,
  K = 5,
  iter_max = 1000,
  burn_in = 500,
  thin = 1,
  fix_alpha = FALSE,
  seed = NULL
)

Arguments

formula	Similar as for sstap_lm, though fdp_staplm is currently restricted to only one stap term.
benvo	built environment object from the rbenvo package containing the relevant data
weights	weights for weighted regression - default is vector of ones
alpha_a	alpha gamma prior hyperparameter or alpha if fix_alpha = TRUE
alpha_b	alpha gamma prior hyperparameter
sigma_a	precision gamma prior hyperparameter
sigma_b	precision gamma prior hyperparameter
tau_a	penalty parameters gamma prior hyperparameter
tau_b	penalty parameters gamma prior hyperparameter
K	truncation number
iter_max	maximum number of iterations
burn_in	number of burn in iterations
thin	number by which to thin samples
fix_alpha	boolean value indicating whether or not to fix the concentration parameter
seed	random number generator seed will be set to default value if not by user

Value

a stapDP model object

Details

This function fits a linear model in a bayesian paradigm with improper priors assigned to the "standard" regression covariates designated in the formula argument and a Dirichlet process prior with normal-gamma base measure assigned to the stap basis function expansion using penalized splines via jagam.

The concentration parameter is assigned gamma prior with hyperparameters shape alpha_a and scale alpha_b. Precision parameters sigma_a,sigma_b, tau_a,tau_b are similar for the residual and penalties' precision, respectively.