I have some data for which, at one level of a factor, there is a significant correlation. At the other level, there is none. Plotting these side-by-side is simple. Adding a l
Don't think about picking a facet, think supplying a subset of your data to stat_smooth:
ggplot(df, aes(x, y)) + geom_point() + geom_smooth(data = subset(df, z =="a")) + facet_wrap(~ z)