trls.influence: Regression diagnostics for trend surfaces

Description

This function provides the basic quantities which are used in forming a variety of diagnostics for checking the quality of regression fits for trend surfaces calculated by surf.ls.

Usage

trls.influence(object)
# S3 method for trls
plot(x, border = "red", col = NA, pch = 4, cex = 0.6,
     add = FALSE, div = 8, …)

Arguments

object, x

Fitted trend surface model from surf.ls

div

scaling factor for influence circle radii in plot.trls

add

add influence plot to existing graphics if TRUE

border, col, pch, cex, …

additional graphical parameters

Value

trls.influence returns a list with components:

raw residuals as given by residuals.trls

hii

diagonal elements of the Hat matrix

stresid

standardised residuals

Cook's statistic

References

Unwin, D. J., Wrigley, N. (1987) Towards a general-theory of control point distribution effects in trend surface models. Computers and Geosciences, 13, 351--355.

Venables, W. N. and Ripley, B. D. (2002) Modern Applied Statistics with S. Fourth edition. Springer.

Examples

Run this code

# NOT RUN {
library(MASS)  # for eqscplot
data(topo, package = "MASS")
topo2 <- surf.ls(2, topo)
infl.topo2 <- trls.influence(topo2)
(cand <- as.data.frame(infl.topo2)[abs(infl.topo2$stresid) > 1.5, ])
cand.xy <- topo[as.integer(rownames(cand)), c("x", "y")]
trsurf <- trmat(topo2, 0, 6.5, 0, 6.5, 50)
eqscplot(trsurf, type = "n")
contour(trsurf, add = TRUE, col = "grey")
plot(topo2, add = TRUE, div = 3)
points(cand.xy, pch = 16, col = "orange")
text(cand.xy, labels = rownames(cand.xy), pos = 4, offset = 0.5)
# }

Run the code above in your browser using DataCamp Workspace