pred.weed {WeedMap} | R Documentation |
Makes joint Bayesian inference and spatial prediction within a model for weed count data and a covariate
pred.weed(nit, thin = 1, x = NULL, xy = NULL, y = NULL, z, wx = NULL, wxy = NULL, i = NULL, alpha = NULL, beta = NULL, lambda = NULL, tau = NULL, kappa = NULL, sd.prop.h = 0.1, sd.prop.alpha = 0.5, sd.prop.beta = 0.5, sd.prop.lambda = 0.1, sd.prop.tau = 0.1, delta.prop.kappa = 2, mprior.alpha, vprior.alpha, mprior.beta, vprior.beta, mprior.kappa, vprior.kappa = 999, mprior.tau = 1, vprior.tau = 999, mprior.lambda = 1, vprior.lambda = 999, n.kappa = 1, kappa.max = 0)
nit |
Number of iterations |
thin |
Thinning of the Markov chain |
x |
Coordinates of sites where weed counts only are given |
xy |
Coordinates of sites where weeds and images are given |
y |
Coordinates of sites where image indexes only are given |
z |
Coordinates of sites where predcited values are sought |
wx |
Weed values at sites x |
wxy |
Weed values at sites xy |
i |
Image values at sites xy and y . Values have to
be concatenated in the same vector in this order |
alpha |
Init value for the shape parameter of w |
beta |
Init value for the scale parameter of w |
lambda |
Init value for the scaling parameter relating w and i as
i = λ w ε |
tau |
Init value for the variance of noise epsilon |
kappa |
Init value for the spatial scale parameter of the Gausian random field |
sd.prop.h |
Standard deviation of the Gaussian increment in the proposal of Gaussian components h |
sd.prop.alpha |
Standard deviation of the Gaussian increment in the
proposal for alpha |
sd.prop.beta |
Standard deviation of the Gaussian increment in the
proposal for beta |
sd.prop.lambda |
Standard deviation of the Gaussian increment in the
proposal for lambda |
sd.prop.tau |
Standard deviation of the Gaussian increment in the
proposal for tau |
delta.prop.kappa |
Maximum number of increments allowed (the
amplitude of an increment being kappa.max /n.kappa ) for a
move in the proposal of kappa |
mprior.alpha |
A priori mean of alpha |
vprior.alpha |
A priori variance of alpha |
mprior.beta |
A priori mean of beta |
vprior.beta |
A priori variance of beta |
mprior.kappa |
A priori mean of kappa |
vprior.kappa |
A priori variance of kappa |
mprior.tau |
A priori mean of tau |
vprior.tau |
A priori variance of tau |
mprior.lambda |
A priori mean of lambda |
vprior.lambda |
A priori variance of lambda |
n.kappa |
Number of steps in the discretisation of the support of
kappa |
kappa.max |
Maximum value in the truncation of the support of kappa |
If standard deviation of the Gaussian increment in the update
of alpha, beta, lambda or tau, or if the step in the
increment of kappa is equal to 0, then this variable is not
processed in the MCMC run and stays at its initial value. This is the
way to specify that inference should not be made on a one or several
variable. See examples in WeedMap
where lambda is
initialised at 1 and not updated (sd.prop.lambda=0
).
If init values are not given, the corresponding parameters are initialised from the prior.
A list whose elements are:
x |
Coordinates of sites where weed counts only are given |
y |
Coordinates of sites where image indexes only are given |
z |
Coordinates of sites where predcited values are sought |
wx |
Weed values at sites x |
i |
Image values at sites xy and y . Values are
concatenated in the same vector in this order |
nit |
Number of iterations |
thin |
Thinning of the Markov chain |
wy.MC |
A matrix with ny rows and nit/thin
columns containing sampled values of wy |
wz.MC |
A matrix with nz rows and nit/thin
columns containing sampled values of wz |
alpha.MC |
A vector of length nit/thin of simulated
alpha values |
beta.MC |
A vector of length nit/thin of simulated
beta values |
lambda.MC |
A vector of length nit/thin of simulated
lambda values |
tau.MC |
A vector of length nit/thin of simulated
tau values |
kappa.MC |
A vector of length nit/thin of simulated
kappa values |
n.kappa |
Number of steps in the discretisation of the support of
kappa |
kappa.max |
Maximum value in the truncation of the support of kappa |
Gilles Guillot
G. Guillot, N. Loren, M. Rudemo, Bayesian spatial prediction of weed intensities from exact count data and picture based indexes, 2006, submitted