Setting colors and size inside aes() and inside geom() in ggplot2

Question

I am a beginner and reading Wickham's ggplot2 book to understand how this thing works. I am a little confused about setting attributes (color, size etc.) inside and outside aes() I have included references I've referred here so far.

I ran XYZ experiments to understand this:

Data:

df <- data.frame(x = c(1,2,3), y = c(1,2,3), z=c("a","b","c"))

Experiment 1:

 ggplot(df, aes(y, y)) +
   geom_point(size = 4, show.legend = TRUE) 

This doesn't show the legend although I've set it to TRUE. I am not sure why this is happening.

Experiment 2:

ggplot(df, aes(y, y)) +
   geom_point(size = 4, aes(colour = z)) #I see the legend

This surprisingly shows the legend. I believe that from references, mapping variables to "color" is a valid reason to show the legend. We have mapped a variable to color.

Experiment 3

ggplot(df, aes(y, y)) +
   geom_point(size = 4, aes(colour = z),color = "red") 

Here, I believe red color is overridden by my second call i.e. color="red". Am I correct?

Experiment 4:

ggplot(df, aes(y, y)) +
   geom_point(size = 4, aes(colour = z,size=20),color = "red") 

I wanted to test whether the size increases if I set size = 20 inside aes. It doesn't. I am not sure why because size = 4 was set before setting size = 20.

Experiment 5:

ggplot(df, aes(y, y)) +
   geom_point(size = 4, color = "red") +
   geom_point(aes(colour = z),size=10) #legend appears, but I have lost border colors. 

Finally, I thought of keeping size = 10 outside of aes because I thought setting size = constant inside aes would be defaulting the value of size. This worked in that the size did increase, but I have lost the legend.

I am completely lost -how do I show the legend, increase the size (using variables or manually) or change colors( using variables or manually)? I understand that changing size and colors would be the easiest because I would just have to map them to variables. However, I am not sure when I would want to keep one of them (say size or color) constant.

 ggplot(df, aes(y, y)) +
   geom_point(aes(colour = z, size = x)) ##this maps color and size to variables.

I'd appreciate any thoughts. Finally, I am sorry if my question is too basic for some of you.

---

References: Difference between passing options in aes() and outside of it in ggplot2

---

Update from the thread: (after discussions for those of us who are reading this...)

Experiment 1: No legend shows up because there is no need for one. There is no variable is mapped to size. There's nothing to show in the legend in that case.

Experiment 2: There is one layer and it maps to the variable. So, I see the legend.

Experiment 3: I have set the color using variable, but then it's overridden by red color.

Experiment 4 Ditto here. The output of Expt 4 and 3 are the same because of overriding issue.

Experiment 5 The second layer overlays on top of first layer. One way to reverse this would be:

ggplot(df, aes(y, y)) + geom_point(aes(colour = z),size=10) +
   geom_point(size = 4, color = "red")

Answer 1

You should read more on defining aesthetics globally, and defining them on local geom_xyz() layers.

The thing is, when you define col as attributes, it overrides your aesthetics (aes()) mapping. You can see that in your Experiment 3.

In the fourth one, you are again overriding the color by mapping it outside of aes().

You should check this link out on aesthetics mapping: http://docs.ggplot2.org/current/aes.html

From the same link:

Aesthetics supplied to ggplot() are used as defaults for every layer

you can override them, or supply different aesthetics for each layer

It's nothing so complicated, you should look at the documentation more carefully and it'll come to you naturally.