spsi (version 0.1)

sps_prep: Prepare list of objects for shape preserving interpolation.

Description

This is primitive function which output is used by sps_eval. Function works with both uni- and bi-variate functions and returns list consisting original data provided, values of the derivatives and degrees of polynomial needed to satisfy the shape constraints required

Usage

sps_prep(x, y, z=NULL, fx = NA, fy = NA, fxy = NA, shape = c("monotonicity", "curvature"), shape.x = shape, shape.y = shape, max.deg = 50, smoothness = 1, tol = 0.0001)

Arguments

x

Vector of x-coordinates of points to be interpolated.

y

Vector of y-coordinates of points to be interpolated. For uni-variate case length should be equal to length of x and $y[i] = f(x[i])$. For bi-variate case length may differ.

z

Bi-variate case only. Matrix of values of function on grid spanned by x and y; $z[i,j] = f(x[i],y[j])$ OR Vector of values of functions on points (x, y); $z[i] = f(x[i],y[i])$. In this case length of all three vectors should be equal and it should be possible to transform the point to a gridded form.

fx

Matrix of values of the derivative with respect to x of the function; $fx[i,j] = fx(x[i],y[j])$; By default it is estimated internally.

fy

Bi-variate case only. Matrix of values of the derivative with respect to y of the function; $fy[i,j] = fy(x[i],y[j])$; By default it is estimated internally.

fxy

Bi-variate case only. Matrix of values of mixed partial derivative of the function; $fxy[i,j] = fxy(x[i],y[j])$; By default it is estimated internally.

shape

Specifies which attributes should be preserved. Vector should contain 'monotonicity' and/or 'curvature' only. In bi-variate case this can be set separately for both dimensions using shape.x, shape.y.

shape.x

Specifies which attributes should be preserved for x dimension.

shape.y

Specifies which attributes should be preserved for y dimension.

max.deg

Specifies maximum degree of polynomial allowed. In some cases in order to preserve shape, very high degrees are necessary. If maximum degree is reached, it is not guaranteed that resulting spline will preserve all the attributes required.

smoothness

How many times does the spline needs to be differentiable.

tol

Tolerance used within program. Default value is suitable for graphical purposes.

Value

Uni-variate: list with 6 components:

x,y

Coordinates of the data points

k

Smoothness (or continuity class) required

fx

Estimated or given values of derivative

deg

Degree of polynomial needed on each of the line segments

dim

Number of variables; numeric equal to 1

Bi-variate: list with 10 components:

x,y

Coordinates of the data points

z

Matrix of vales of the function

k

Smoothness (or continuity class) required

fx, fy, fxy

Estimated or given values of respective derivatives

deg.x, deg.y

Degree of polynomial needed on each of the line segments in each dimension

dim

Number of variables; numeric equal to 2

Details

If z is not provided function will prepare list needed for uni-variate interpolation. If values of the derivatives are provided, resulting spline will preserve them. For bi-variate case it is possible to set some derivatives and let program estimate the rest.

References

Costantini, P; Fontanella, F; 'Shape Preserving Bi-variate Interpolation' SIAM J NUMER. ANAL. Vol. 27, No.2, pp. 488-506, April 1990

See Also

sps_eval sps_fun

Examples

## Univariate example
x <- c( 1, 2, 3, 4, 5, 6)
y <- c(16 ,18, 21, 17, 15, 12)
spline <- sps_prep(x, y, shape = 'monotonicity', smoothness = 2)
plot(seq(1, 6, 0.1), sps_eval(spline, seq(1, 6, 0.1)))

## Bivariate example

tower <- function(x, y)
{
	X <- abs(x)
	Y <- abs(y)
	ifelse((X + Y) <= 1, floor(3*(1 - X - Y)),
	       ifelse(pmax(X, Y) >= 1, pmax(X, Y)/2 - 0.5,
                      0))
}

X <- Y <- seq(-1.25, 1.25, 2.5/13)
grid <- mesh(X, Y)
Z <- tower(grid$x, grid$y)

spline <- sps_prep(X, Y, Z)

X_ <- Y_ <- seq(-1.25, 1.25, 2.5/60)

persp3D(X_, Y_, sps_eval(spline, x = X_, y = Y_, grid = TRUE))