19

I do the following graph:

> ddd
      UV.NF TRIS            volvol
2  145.1923   31  500 µl / 625  µl
3  116.3462   50  500 µl / 625  µl
4  127.1635   60  500 µl / 625  µl
5  125.9615   69  500 µl / 625  µl
6  162.0192   30    1 ml / 625  µl
7  166.8269   50    1 ml / 625  µl
8  176.4423   60    1 ml / 625  µl
9  171.6346   70    1 ml / 625  µl
19 292.3077   31 500 µl / 2500  µl
20 321.1538   50 500 µl / 2500  µl
21 225.0000   60 500 µl / 2500  µl
22 263.4615   69 500 µl / 2500  µl
23 301.9231   30   1 ml / 2500  µl
24 350.0000   50   1 ml / 2500  µl
25 282.6923   60   1 ml / 2500  µl
26 282.6923   70   1 ml / 2500  µl
35 133.6207   31  500 µl / 625  µl

ggplot() +  
    geom_point(aes(y = log(UV.NF), x = TRIS, colour=ddd[,"volvol"], shape=ddd[,"volvol"]), 
        data=ddd) + 
    labs(colour = "volvol", shape="volvol") + xlab("TRIS (mM)") + 
    guides(colour = guide_legend(title="Vol. lyo. / Vol. reconst."), 
        shape=guide_legend(title="Vol. lyo. / Vol. reconst.")) +
    scale_shape_manual(values = c(19,19,3,3)) + scale_colour_manual(values = c(2,4,2,4))

graph

I want to add the regression line lm(y~x) for each of the four groups appearing in the legend. I have done many attempts with geom_smooth() but without success.

Stéphane Laurent
  • 75,186
  • 15
  • 119
  • 225

1 Answers1

30

I'm not quite sure whether that's what you want, but have you tried the following?

ggplot(ddd,aes(y = log(UV.NF), x = TRIS, colour = volvol, shape = volvol)) +
   geom_point() + geom_smooth(method = "lm", fill = NA)

This gives me the following plot with your data:enter image description here

There's also some documentation for geom_smooth that does pretty much what you'd like, albeit in a more complicated (yet flexible) manner.

Henrik
  • 65,555
  • 14
  • 143
  • 159
Aleksandar Dimitrov
  • 9,275
  • 3
  • 41
  • 48
  • Note that with so few data points, the linear model isn't very precise. – Aleksandar Dimitrov Sep 05 '12 at 12:46
  • 2
    Excellent ! That did not work because I typed ggplot() + geom_point(...) instead of ggplot(...) + geom_point() – Stéphane Laurent Sep 05 '12 at 12:49
  • 3
    You should *always* put your data and general aesthetics (aes()) in the ggplot function, except you've a very good reason not to (when you want to put different kinds of plots on the same graph, then it makes sense to put the aes() in the respective geom_* functions. – Aleksandar Dimitrov Sep 05 '12 at 12:51