predict.bsam: Predict method for a bsam object

Description

Computes the predicted values of Bayesian spectral analysis models.

Usage

# S3 method for bsam
predict(object, newp, newnp, alpha = 0.05, HPD = TRUE, type = "response", ...)

Value

A list object of class predict.bsam containing posterior means and \(100(1-\alpha)\)% credible intervals.

The output list includes the following objects:

fxobs: posterior estimates for unknown functions over observation.
wbeta: posterior estimates for parametric part.
yhat: posterior estimates for fitted values of either response or expectation of response. For gbsar, it gives posterior estimates for expectation of response.
fxResid: posterior estimates for fitted parametric residuals. Not applicable for gbsar.

Arguments

object: a bsam object
newp: an optional data of parametric components with which to predict. If omitted, the fitted values are returned.
newnp: an optional data of nonparametric components with which to predict. If omitted, the fitted values are returned.
alpha: a numeric scalar in the interval (0,1) giving the \(100(1-\alpha)\)% credible intervals.
HPD: a logical variable indicating whether the \(100(1-\alpha)\)% Highest Posterior Density (HPD) intervals are calculated. If HPD=FALSE, the \(100(1-\alpha)\)% equal-tail credible intervals are calculated. The default is TRUE.
type: the type of prediction required. type = "response" gives the posterior predictive samples as default. The "mean" option returns expectation of the posterior estimates.
...: not used

Details

None.

Examples

Run this code

if (FALSE) {

##########################################
# Increasing Convex to Concave (S-shape) #
##########################################

# simulate data
f <- function(x) 5*exp(-10*(x - 1)^4) + 5*x^2

set.seed(1)

n <- 100
x <- runif(n)
y <- f(x) + rnorm(n, sd = 1)

# Number of cosine basis functions
nbasis <- 50

# Fit the model with default priors and mcmc parameters
fout <- bsar(y ~ fs(x), nbasis = nbasis, shape = 'IncreasingConvex',
             spm.adequacy = TRUE)

# Prediction
xnew <- runif(n)
predict(fout, newnp = xnew)
}