BCfit {BayesComm} | R Documentation |
BCfit
is the workhorse function for the BayesComm model.
It is highly recommended to use the wrapper function BC
which checks inputs and sets up different model types and initial values.
BCfit
arguments can be accessed through BC
using the ...
argument.
BCfit(y, X, covlist, R, z, mu, updateR, iters, thin = 1, burn = 0, priW = c(nrow(z) + 2 * ncol(z), 2 * ncol(z)), verbose = 0)
y |
matrix of species presence/absence data |
X |
matrix of environmental covariates |
covlist |
optional list of which covariates to assign to each species |
R |
initial values for correlation matrix |
z |
initial values for z |
mu |
initial values for mu |
updateR |
logical; if true the correlation matrix is updated, if false it is fixed at |
iters |
total number of iterations |
thin |
amount to thin the posterior chains. Defaults to 1 (no thinning) |
burn |
number of iterations to discard at the beginning of the chain |
priW |
prior specification for correlation matrix W |
verbose |
how often to print updates to the console. |
priW
specifies the inverse Wishart prior on the unknown and unidentifiable covariance matrix W from which the correlation matrix R is derived.
priW
is a vector of length two, the first element specifies the degrees of freedom, the second element is multiplied by an identity matrix to form the scale matrix.
The default for priW
is c(n + 2p, 2p), where n is the number of records and p is the number of species in the community; this therefore forms the prior: iW(n + 2p, 2pI).
This prior was determined to exert minimal influence on the posterior of R whilst limiting dependence of R on the unidentifiable variance parameters of W.
For further details on how to specify Y
, X
and covlist
see BC
.
A list containing elements:
R |
samples from posteriors of the correlation matrix |
B |
samples from posteriors of regression coefficients (a list of matrices) |
z |
samples from posteriors of latent variables z |