makeBasis: Construction of quantile-based linear spline basis

In hypergsplines: Bayesian model selection with penalised splines and hyper-g prior

Description

Construction of quantile-based linear spline basis

Usage

  makeBasis(x, type = c("linear", "cubic"), nKnots,
    settings = NULL)

Arguments

x	the covariate vector
type	the type of splines to be used, “linear” splines modelled by a truncated power series basis, or “cubic” splines modelled by O'Sullivan splines.
nKnots	number of quantile-based knots to be used
settings	list of settings. By default, this is NULL and a whole new basis depending on type and nKnots is built. Otherwise this is expected to be attributes(basis) and the settings of the old basis are used to compute the spline basis vectors at new grid points x.

Value

the spline basis in a matrix with length(x) rows. As an attribute, a list with the settings is provided.

Author(s)

Daniel Sabanes Bove [e-mail]

Examples

x <- seq(from=-5,
         to=5,
         length=1000)

Z <- makeBasis(x=x,
               nKnots=4L,
               type="linear")
str(Z)

plot(x,
     Z %*% c(1, -3, 2, 1),
     type="l")
abline(v=attr(Z, "knots"),
       col="gray",
       lty=2)

Z <- makeBasis(x=x,
               nKnots=10L,
               type="cubic")
dimZ <- ncol(Z)

coefs <- rnorm(dimZ)

plot(x,
     Z %*% coefs,
     type="l")
abline(v=attr(Z, "knots"),
       col="gray",
       lty=2)

hypergsplines documentation built on May 2, 2019, 6:14 p.m.