bkmrdlm | R Documentation |
This estimates the Bayesian kernel machine regression - distributed lag model (BKMR-DLM).
bkmrdlm(
y,
x,
z,
niter,
nburn = round(niter/2),
nthin = 1,
prior_sigma = c(0.1, 0.1),
prior_tau = 0.1,
kappa = 1,
basis.opts = list(type = "gam", pve = 0.9),
gaussian = TRUE,
polydegree = 2
)
y |
Vector of outcomes. |
x |
A list of matricies. Each Matric should be n by T with each row being the T-vector of exposures for one individuals. When only one exposure is used x can be a matrix instead of a list. |
z |
a matrix or design frame of covaiates and confounds. This can be ommited. |
niter |
Number of MCMC iterations including burnin. |
nburn |
The number of iterations to be discarded as burnin. |
nthin |
Thining, every nthin-th draw from the posterior will be saved. |
prior_sigma |
Vector of length 2 with the parameters for the gamma prior on sigma^-2 |
prior_tau |
the prior variance on log(tau^2). |
kappa |
scale parameter, rho/kappa~chisq(1). |
basis.opts |
List with the entries: type = the type of basis used, either 'face' (default) or "ns" or "bs" for splines or "gam" for presmoothing the exposure with a gam following defaults from mgcv; knots = the number of knots used for method face; pve = the percent of variance explained by the PCs for method face; df = the df for ns method. |
gaussian |
Use a Gaussian kernel (TRUE, default) or a polynomial kernel (FALSE) |
polydegree |
Degree of polynomial when polynomial kernel is used. Only applies when gaussian=FALSE. |
An object of class 'bkmrdlm'.
Ander Wilson
library(regimes)
#simulate data from scenario A
dat <- simBKMRDLM(n = 200, scenario="A", sd=1, seed=1234)
# Estimate model
# This may take a few minutes
# Increase iterations for a real analysis
fit <- bkmrdlm(y=dat$y,
x=dat$x,
z=dat$z,
niter=100,
gaussian=FALSE,
polydegree=2)
summary(fit)
plot(fit)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.