spreg {lazy.stat}

R Documentation

Spline Regression Analysis: Regress y on spline expanded univariate regressor variable x.

Description

Spline Regression Analysis: Regress y on spline expanded univariate regressor variable x.

Usage

spreg(
  y,
  x,
  degree = 3,
  knots = NULL,
  n = 0,
  nknots = n,
  intercept = 1,
  xmin = NULL,
  xmax = NULL,
  type = "b",
  maxiter = 50,
  eps = 1e-06,
  epsg = 1e-06,
  MP = 1,
  epsSwp = 1e-08,
  nopermute = 0,
  rescale = 0,
  orth = 0,
  xrange = NULL,
  plotbase = 0,
  printbase = 0,
  print = 0,
  xlim = NULL,
  ylim = NULL,
  title = NULL,
  xlab = "new",
  plot = 0
)

Arguments

y	Input criterion variable vector
x	Input regressor vector.
degree	Degree of polynomial to be used.
knots	vector consisting of (interior) knots to be used.
n	# of knots to be used in [xmin, xmax]. Same as nknots
nknots	# of knots to be used in [xmin, xmax].
intercept	= 0 to omit the intercept when type="p".
xmin	minimum of interior knots.
xmax	maximum of interior knots.
type	= "p" for piecewise spline, = "b" for b-spline = "m" to perform monotone spline regression.
maxiter	Max # of iterations for monotone spline
eps	Criterion for relative improvement of rss for monotone spline
epsg	Criterion for max(abs(gradient)) of rss for monotone spline
MP	Type of ginv to be used
epsSwp	Criterion to judge if the pivot is zero in matSwp
nopermute	= 0 not to permute rows/cols in matSwp
rescale	= 1 to rescale all the p-spline basis in [0,1].
orth	= 1 to orthogonalize the p-spline basis.
xrange	= new range of x or NULL
plotbase	= 1 to plot the spline basis matrix.
printbase	= 1 to print the spline basis matrix.
print	= 1 to print the result
xlim	xlim for plot.
ylim	ylim for plot.
title	Plot title.
xlab	="org" to use old value of x when p-spline with xrange.
plot	= 1 to plot the result

Details

When knots are present, it has priority over nknots.

When type="m", b-spline base is be created and the regression coefficient beta will be estimated with the constraints:
beta[1] <= beta[2] <= .... <= beta[ncol(X)]

Value

List of: x
y
type
rescale, orth, reflect
xrange, xrange0, Xrange0, knots_org
beta
yhat
X=length(x) x (nknots+degree+1) spline basis matrix
knots The interior knots
boundary.knots The boundary knots for bspline

Examples

# artificial data
set.seed(1701)
x <- seq( 0, 3, length=50 )
yhat <- -sin(1*pi*x^(1/2))^3
y <- yhat+0.1*rnorm(length(x))

# data plot
plot( x, y, type="p", ylim=c(-1.5,1.5), main="Data and True Func" )
lines( x, yhat )

# Simple Spline Regression
res=spreg( y, x, n=5, plot=1, ylim=c(-1.5,1.5) )

# Monotore Spline Regression
res=spreg( y, x, n=5, plot=1, ylim=c(-1.5,1.5), type="m" )

# Monotore LS Regression
resm=mlsreg( y,x, method=3, print=1, plot=1 )

---