mgcv (version 1.8-31)

place.knots: Automatically place a set of knots evenly through covariate values

Description

Given a univariate array of covariate values, places a set of knots for a regression spline evenly through the covariate values.

Usage

place.knots(x,nk)

Arguments

x

array of covariate values (need not be sorted).

nk

integer indicating the required number of knots.

Value

An array of knot locations.

Details

Places knots evenly throughout a set of covariates. For example, if you had 11 covariate values and wanted 6 knots then a knot would be placed at the first (sorted) covariate value and every second (sorted) value thereafter. With less convenient numbers of data and knots the knots are placed within intervals between data in order to achieve even coverage, where even means having approximately the same number of data between each pair of knots.

References

http://www.maths.bris.ac.uk/~sw15190/

See Also

smooth.construct.cc.smooth.spec

Examples

# NOT RUN {
require(mgcv)
x<-runif(30)
place.knots(x,7)
rm(x)
# }