CODEWITHSUNDEEP
rggplot2anovainteraction
blazej
blazej

Reputation: 1788

4-way interaction with ggplot2 lines and points defined by color and linetype

I'm trying to plot a 4-way interaction from a factorial experiment using ggplot2 with geom_line() and geom_point().

Data I'm working with is emmeans() object with marginal means estimated from a linear mixed-effects model.

And the graph I can get via emmip() is this:

enter image description here

What I'd like is to separate b:c interaction so that one factor is defined by color and / or point type and the other by linetype (dashed vs. solid)

A minimal example is this:

df<- data.frame(y=rnorm(n=16),
                a=gl(2,4,16, labels=c("a1","a2")),
                b=gl(2,2,16, labels=c("b1", "b2")),
                c=gl(2,1,16, labels=c("c1", "c2")),
                fac=gl(2,8,16, labels=c("panel1", "panel2")))

I tried the following ggplot() code:

ggplot(df, aes(y=y, x=a, color=b)) + 
      geom_point(aes(shape=b), size=3) +
      geom_line(aes(linetype=c)) +
      facet_wrap(~fac)

But I get a warning that I don't know how to handle:

geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic?
geom_path: Each group consists of only one observation. Do you need to adjust the group aesthetic?

And the graph looks almost as intended, but without the c factor horizontal lines drawn between corresponding b points:

enter image description here

How can I fix this?

I found a similar question dealing directly with predicted values from a lmer() object: How to plot mixed-effects model estimates in ggplot2 in R? but still can't find a way to deal with my data format.

Upvotes: 3

Views: 3768

Answers (2)

blazej
blazej

Reputation: 1788

Following Z.Lin comment above I found what I was looking for with:

ggplot(df, aes(y=y, x=a)) + 
      geom_point(aes(color=b),size=3) +
      geom_line(aes(group = b:c, color=b, linetype=c), size=1) +
      facet_wrap(~fac)

enter image description here

Upvotes: 3

ozanstats
ozanstats

Reputation: 2864

One problem with your example is that you use both color and shape aesthetics for the variable b. I would recommend to choose one and save the other for the variable c rather than using lines. Generally speaking, using lines makes sense for continuous variables and trends. You can use one of the following:

# colored by b & shaped by c
ggplot(df, aes(y=y, x=a, color=b)) + 
  geom_point(aes(shape=c), size=3) +
  facet_wrap(~fac)

# colored by c & shaped by b
ggplot(df, aes(y=y, x=a, color=c)) + 
  geom_point(aes(shape=b), size=3) +
  facet_wrap(~fac)

Upvotes: 0

Related Questions