Plotting xyplot with regression line on lattice graphics

Question

I have loaded the lattice package. Then I run:

> xyplot(yyy ~ xxx | zzz, panel = function(x,y) { panel.lmline(x,y)}

This produces panel

Answer 1

All you really need is:

xyplot(yyy ~ xxx | zzz, type = c("p","r"))

where the type argument is documented in ?panel.xyplot

I won't quote it all but

type: character vector consisting of one or more of the following:
      ‘"p"’, ‘"l"’, ‘"h"’, ‘"b"’, ‘"o"’, ‘"s"’, ‘"S"’, ‘"r"’,
      ‘"a"’, ‘"g"’, ‘"smooth"’, and ‘"spline"’.  If ‘type’ has more
      than one element, an attempt is made to combine the effect of
      each of the components.

      The behaviour if any of the first six are included in ‘type’
      is similar to the effect of ‘type’ in ‘plot’ (type ‘"b"’ is
      actually the same as ‘"o"’).  ‘"r"’ adds a linear regression
      line (same as ‘panel.lmline’, except for default graphical
      parameters). ‘"smooth"’ adds a loess fit (same as
      ‘panel.loess’).  ‘"spline"’ adds a cubic smoothing spline fit
      (same as ‘panel.spline’).  ‘"g"’ adds a reference grid using
      ‘panel.grid’ in the background (but using the ‘grid’ argument
      is now the preferred way to do so).  ‘"a"’ has the effect of
      calling ‘panel.average’, which can be useful for creating
      interaction plots.  The effect of several of these
      specifications depend on the value of ‘horizontal’.

You can, as I showed above, add these in series by passing type a character vector. Essentially, your code gave the same result as type = "r", i.e. only the regression line was drawn.

The panel argument of xyplot, and Lattice plotting functions in general, is extremely powerful but not always required to so quite complex things. Basically you need to pass panel a function which will draw things on each panel of the plot. To modify your code to do what you wanted, we need to add a call to panel.xyplot() as well. E.g.:

xyplot(yyy ~ xxx | zzz,
       panel = function(x, y, ...) {
                 panel.xyplot(x, y, ...)
                 panel.lmline(x, y, ...)
               })

It is also very useful to pass all other arguments on the individual panel functions via ..., in which case you need ... as an argument in your anonymous functions (as shown above). In fact, you can probably write that panel function part as:

xyplot(yyy ~ xxx | zzz,
       panel = function(...) {
                 panel.xyplot(...)
                 panel.lmline(...)
               })

but I usually add the x and y arguments just to be clear.