Ggplot2 - How to change the geometry order in a legend only?

Question

I am doing a plot in ggplot2 that contains geom_pointrange and geom_line. I see that when I change the order of the geometries, either the points are built on top of the line, or vice versa. The legend also changes geometry from above to another, based on the same geometry order. However, I would like the line to be drawn first, then the pointer above, in the plot itself, with the opposite in the legend. Is it possible? I would really appreciate any material.

Here is the code I used to create the drawing.

md.figd2 <- structure(list(date = c("2013-05-28", "2013-07-11", "2013-09-22", "2013-05-28", "2013-07-11", "2013-09-22", "2013-05-28", "2013-07-11", "2013-09-22"), trt = structure(c(3L, 3L, 3L, 1L, 1L, 1L, 2L, 2L, 2L), .Label = c("- Fescue", "- Random", "Control"), class = "factor"), means = c(1, 0.921865257043089, 0.793438250521971, 1, 0.878305313846414, 0.85698797555687, 1, 0.840679145697309, 0.798547331410388 ), mins = c(1, 0.87709562979756, 0.72278951032918, 1, 0.816185624483356, 0.763720265496049, 1, 0.780804129401513, 0.717089626439849 ), maxes = c(1, 0.966634884288619, 0.864086990714762, 1, 0.940425003209472, 0.950255685617691, 1, 0.900554161993105, 0.880005036380927)), .Names = c("date", "trt", "means", "mins", "maxes"), row.names = c(NA, 9L), class = "data.frame") library(ggplot2) dplot1.ysc <- scale_y_continuous(limits=c(0,1), breaks=seq(0,1,.2), name='Proportion mass lost') dplot1.xsc <- scale_x_date(limits=as.Date(c('2013-05-23', '2013-10-03')), labels=c('May 28', 'July 11', 'Sep 22'), breaks=md.figdata$date, name='Date') dplot1.csc <- scale_color_manual(values=c('grey20','grey50','grey80')) dplot1.lsc <- scale_linetype_manual(values=c('solid','dotted','dashed')) djitter <- rep(c(0,-1,1), each=3) # This one produces the plot with the legend I want. dplot1b <- ggplot(md.figd2, aes(x=date + djitter, y=means, group=trt)) + geom_pointrange(aes(ymin=mins, ymax=maxes, color=trt), size=2) + geom_line(aes(linetype=trt), size=1) # This one produces the plot with the points on the main plot that I want. dplot1b <- ggplot(md.figd2, aes(x=date + djitter, y=means, group=trt)) + geom_line(aes(linetype=trt), size=1) + geom_pointrange(aes(ymin=mins, ymax=maxes, color=trt), size=2) dplot1b + dplot1.xsc + dplot1.ysc + dplot1.csc + dplot1.lsc 

Answer 1

Quick and dirty: add another geom_line that uses data outside ylim and place it after geom_pointrange in the order of plotting.

 # some data for the 'fake' line df <- data.frame(date = rep(as.Date(c('2013-05-23', '2013-10-03')), 3), means = 1.5, trt = c("- Fescue", "- Random", "Control")) # insert fake line after geom_pointrange ggplot(md.figd2, aes(x = date, y = means, group = trt)) + geom_line(aes(linetype = trt), size = 1) + geom_pointrange(aes(ymin = mins, ymax = maxes, color = trt), position = position_dodge(width = 5), size = 2) + geom_line(data = df, aes(linetype = trt), size = 1) + scale_y_continuous(limits = c(0,1), breaks = seq(0,1, 0.2), name = 'Proportion mass lost') + scale_x_date(limits = as.Date(c('2013-05-23', '2013-10-03')), name = 'Date') + scale_color_manual(values = c('grey20','grey50','grey80')) + scale_linetype_manual(values = c('solid','dotted','dashed')) 

Note that I deleted some of your chart settings in my example, in part because object 'md.figdata' not found .

Answer 2

You can use gtable::gtable_filter to extract the legend from the desired graph, and then gridExtra::grid.arrange to recreate the plot you want

 # the legend I want plot1a <- ggplot(md.figd2, aes(x=date , y=means, group=trt)) + geom_pointrange(aes(ymin=mins, ymax=maxes, color=trt), size=2, position = position_dodge(width=1)) + geom_line(aes(linetype=trt), size=1) # This one produces the plot with the points on the main plot that I want. dplot1b <- ggplot(md.figd2, aes(x=date, y=means, group=trt)) + geom_line(aes(linetype=trt), size=1) + geom_pointrange(aes(ymin=mins, ymax=maxes, color=trt), size=2) w <- dplot1b + dplot1.xsc + dplot1.ysc + dplot1.csc + dplot1.lsc # legend l <- dplot1a + dplot1.xsc + dplot1.ysc + dplot1.csc + dplot1.lsc library(gtable) library(gridExtra) # extract legend ("guide-box" element) leg <- gtable_filter(ggplot_gtable(ggplot_build(l)), 'guide-box') # plot the two components, adjusting the widths as you see fit. grid.arrange(w + theme(legend.position='none'),leg,ncol=2, widths = c(3,1)) 

An alternative is to simply replace the legend in the plot you want with the legend you want to extract (using gtable_filter )

 # create ggplotGrob of plot you want wGrob <- ggplotGrob(w) # replace the legend wGrob$grobs[wGrob$layout$name == "guide-box"][[1]] <- leg grid.draw(wGrob)