Question

I am trying to replicate the example, given in Statistical Models in S. Here is the code given in the book.

data(ethanol)
attach(ethanol)
E.intervals <- co.intervals(E, number = 9, overlap = 1/4)
coplot(NOx ~ C | E, given.values = E.intervals,
                    panel = function(x,y) panel.smooth(x,y,degree = 1, span = 1))

This code works only in S, because panel.smooth function doesn’t have degree argument.
So I used

coplot(NOx ~ C | E, given.values = E.intervals,
                    panel = function(x,y,...) panel.smooth(x,y, iter = 1, span = 1))

I would like to use lattice graphics package because it has panel.loess function that is similar to panel.smooth from S, but I haven’t found the function similar to coplot. I think it is xyplot. If it is, can you suggest me the parameter equivalent to given.values?

Answer 1

Here is how it might be done, it is not quite the same but very close.

require(gridExtra)
require(lattice)
data(ethanol)
E.intervals <- equal.count(E, number = 9, overlap = 1/4)
plot1 <- plot(E.intervals)
plot2 <- xyplot(NOx ~ C | E.intervals, data = ethanol, span = 1, layout = c(3,3),
                panel = function(x,y,span){
                panel.xyplot(x, y);
                panel.loess(x, y, span);
               })
grid.arrange(plot1,plot2, nrow=2)

The difference is the hight of the upper plot, it should be smaller.