lpc.spline {LPCM}

Representing local principal curves through a cubic spline.

Description

Fits a natural cubic spline component-wise through the series of local centers of mass. This provides a continuous parametrization in terms of arc length distance, which can be used to compute a projection index for the original or new data points.

Usage

lpc.spline(lpcobject, optimize = TRUE, compute.Rc=FALSE,
     project=FALSE, ...)

Arguments

lpcobject

Object of class lpc.

optimize

Boolean. If TRUE, optimize is used to find the point on the curve with minimum distance. Otherwise, data points are only projected onto the closest knot.

compute.Rc

Boolean. If TRUE, the goodness-of-fit measure R_C suggested in [1] is computed and returned (using the scaled data, if scaled=TRUE in lpcobject).

project

Boolean. If TRUE, projections onto curve are computed.

...

Additional arguments to be passed to lpc.project.spline

Details

See reference [2].

Value

knots.pi

LPC parameters (in cubic spline parametrization) at position of the knots of the spline function (these are not identical to the LPC mass points!)

knots.coords

Coordinates of the spline knots.

closest.pi

Parameter of the projected data points.

closest.coords

Coordinates of projected data points.

closest.dist

Euclidean distance between original and projected data point.

closest.branch

ID Number of the branch on which the data point was projected (the IDs are given in the output of function lpc).

Rc

Value of R_C.

project

repeats the input value of project.

lpcobject

returns the provided object lpcobject.

splinefun

returns the cubic spline function (generated by lpc.splinefun).

Warning

Careful with options project and compute.Rc - they can take rather long if the data set is large!

Note

The parametrization of the cubic spline function is not exactly the same as that of the original LPC. The reason is that the latter uses Euclidean distances between centers of masses, while the former uses the arc length along the cubic spline. However, the differences are normally quite small.

Author(s)

J. Einbeck and L. Evers

References

[1] Einbeck, J., Tutz, G., and Evers, L. (2005). Local principal curves. Statistics and Computing 15, 301-313.

[2] Einbeck, J., Evers, L. & Hinchliff, K. (2010): Data compression and regression based on local principal curves. In A. Fink, B. Lausen, W. Seidel, and A. Ultsch (Eds), Advances in Data Analysis, Data Handling, and Business Intelligence, Heidelberg, pp. 701–712, Springer.

See Also

lpc

Examples

data(gvessel)
gvessel.lpc <- lpc(gvessel[,c(2,4,5)],   h=0.11,  x0=c(35, 1870, 6.3))
gvessel.spline  <- lpc.spline(gvessel.lpc)
plot(gvessel.spline, lwd=2)