Data visualization in R and concepts of ggplot2:

See my presentation slides here (right-click and open the link in a new tab).

Keep the concepts for ggplot2 in mind: layers of data + aesthetic mappings + geometries

Exploratory graphics with base R

Load some data (simplified Ellis & Yuan 2004 data and manipulated Obarow from Larson-Hall, 2015)

ell <- read.csv("EllisYuan.csv")
obarow <- read.csv("obarow.csv")
str(obarow)
## 'data.frame':    67 obs. of  10 variables:
##  $ id       : int  3 2 6 7 1 10 5 8 11 14 ...
##  $ gender   : Factor w/ 2 levels "female","male": 2 2 2 2 2 1 2 2 1 1 ...
##  $ grade    : int  1 1 1 1 1 1 2 2 2 2 ...
##  $ treatment: Factor w/ 4 levels "NMNP","NMYP",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ pretest  : int  15 11 13 14 13 14 18 16 15 17 ...
##  $ posttest : int  14 11 13 15 12 14 16 14 13 16 ...
##  $ gain1    : int  -1 0 0 1 -1 0 -2 -2 -2 -1 ...
##  $ gain2    : int  -2 1 0 1 0 5 4 2 1 -1 ...
##  $ gain3    : int  0 0 1 3 -1 -1 0 0 1 1 ...
##  $ gain4    : int  0 3 1 0 1 -1 0 0 0 1 ...
plot(posttest ~ treatment, data = obarow)

plot(posttest ~ pretest, data = obarow)

plot(treatment ~ gender, data = obarow)

The plot function will automatically generates plots suitable for the data type

Let’s use ggplot2: Load the library

library(ggplot2)

Example 1: Graphics for Comparing Groups

STEP 1: Data : Decide the layout

  1. Inspect the data structure and select variables
str(ell)
## 'data.frame':    60 obs. of  2 variables:
##  $ condition: Factor w/ 3 levels "NP","OLP","PTP": 1 1 1 1 1 1 1 1 1 1 ...
##  $ variety  : int  20 15 12 18 9 13 14 14 25 21 ...
  • Dependent/Outcome/Response variable: variety
  • Independent/Predictor variable: condition
  1. Create coordinates with aesthetics mapping (e.g., x-axis, y-axis)
ggplot(data = ell, mapping = aes(x = condition, y = variety))

STEP 2: Elements: Decide which visual elements to include

  1. Add a geometric layer, boxplot, using geom_boxplot function
ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot()

  1. Add data points using geom_point:
ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot() + 
    geom_point()

  1. Add statistical information: Means and CIs with stat_summary We tell R to create additional data, the means and 95% CIs (i.e., “mean_cl_normal”) and represent this new data in the form of a “pointrange” (i.e., mean as a dot and CIs as extending lines).
ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot() + 
    stat_summary(fun.data = "mean_cl_normal", # the data to plot are mean and CIs
                 geom = "pointrange") # use the pointrange shape

  1. Put everything together:
ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot() + 
    geom_point() + 
    stat_summary(fun.data = "mean_cl_normal", geom = "pointrange")

STEP 3: Attributes: Edit attributes

Once you decide the layout and elements of your plot, you can manipuate attributes to increase the readability and explanatory power.

Option 1: boxplot with jittered data points 1-1) Jitter points with position

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot() + 
    geom_point(position = position_jitter()) + 
    stat_summary(fun.data = "mean_cl_normal", geom = "pointrange")

An alternative:

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot() + 
    geom_jitter() + 
    stat_summary(fun.data = "mean_cl_normal", geom = "pointrange")

1-2) Change the width, alpha level, size

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(outlier.size = -1) + # don't show the outliers here
    geom_point(position = position_jitter(width = .2), alpha = .5, size = 2) + 
    stat_summary(fun.data = "mean_cl_normal", geom = "pointrange")

2-1) Modify stat_sumamry: Change the color

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(outlier.size = -1) + 
    geom_point(position = position_jitter(width = .2), alpha = .5) + 
    stat_summary(fun.data = "mean_cl_normal", geom = "pointrange", 
                 color = "firebrick")

2-2) Nudge position

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(outlier.size = -1) + 
    geom_point(position = position_jitter(width = .2), alpha = .5) + 
    stat_summary(fun.data = "mean_cl_normal", geom = "pointrange", 
                 color = "firebrick", position = position_nudge(x = .4))

Option 2: boxplot with dotplot

  1. Add data points using geom_dotplot
ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(outlier.size = -1) + 
    geom_dotplot(binaxis = "y", stackdir = "center", binwidth = .5, alpha = .5)

  • binaxis and stackdir always have to be set like this when used with boxplots
  • modify the bindiwdth to adjust the size of the dots

2-1) Add statistical information: Mean

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(width = .5, outlier.size = -1) + 
    geom_dotplot(binwidth = .5, binaxis = "y", stackdir = "center", alpha = .5) + 
    stat_summary(fun.y = "mean", geom = "point", shape = 18, 
                 color = "firebrick", size = 3)

2-2) Add statistical information: CIs in errorbars

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(width = .5, outlier.size = -1) + 
    geom_dotplot(binwidth = .5, binaxis = "y", stackdir = "center", alpha = .5) + 
    stat_summary(fun.y = "mean", geom = "point", shape = 18, size = 5, 
                 color = "firebrick", position = position_nudge(x = .1)) + 
    stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = .1, 
                 color = "firebrick", position = position_nudge(x = .1))

  1. Edit scale: x-axis and y-axis
ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(width = .5, outlier.size = -1) + 
    geom_dotplot(binwidth = .8, binaxis = "y", stackdir = "center", alpha = .5) + 
    stat_summary(fun.y = "mean", geom = "point", shape = 18, size = 5, 
                 color = "firebrick", position = position_nudge(x = .1)) +
    stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = .1, 
                 color = "firebrick", position = position_nudge(x = .1)) + 
    # change the x-axis text
    scale_x_discrete(labels = c("NP\n(n = 19)", "OLP\n(n = 20)", "PTP\n(n = 21)")) + 
    # change the y-axis scale
    scale_y_continuous(limits = c(0, 35), expand = c(0, 0))

  • scale_x_discrete() has to do with the x-axis and works when the variable is categorical
  • scale_y_continuous() works with y-axis that is continuous. limits controls the y-axis limits, setting expand to c(0, 0) removes paddings that appear top and bottom

STEP 4: Labels

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(width = .5, outlier.size = -1) + 
    geom_dotplot(binwidth = .8, binaxis = "y", stackdir = "center", alpha = .5) + 
    stat_summary(fun.y = "mean", geom = "point", shape = 18, size = 5, 
                 color = "firebrick", position = position_nudge(x = .1)) +
    stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = .1, 
                 color = "firebrick", position = position_nudge(x = .1)) + 
    scale_x_discrete(labels = c("NP\n(n = 19)", "OLP\n(n = 20)", "PTP\n(n = 21)")) + 
    scale_y_continuous(limits = c(0, 35), expand = c(0, 0)) + 
    # title, subtitle, axis labels, legend title
    labs(title = "The Effect of Planning Time on the Amount of Syntactic Variety", 
         subtitle = "in three different conditions", 
     x = "", y = "Syntactic Variety\n")

STEP 5: Theme

ggplot(data = ell, mapping = aes(x = condition, y = variety)) + 
    geom_boxplot(width = .5, outlier.size = -1) + 
    geom_dotplot(binwidth = 1, binaxis = "y", stackdir = "center", alpha = .5) + 
    stat_summary(fun.data = "mean_cl_normal", geom = "errorbar", width = .1, 
                 color = "firebrick", position = position_nudge(x = .2)) + 
    stat_summary(fun.y = "mean", geom = "point", shape = 18, size = 5, 
                 color = "firebrick", position = position_nudge(x = .2)) + 
    scale_x_discrete(labels = c("NP\n(n = 19)", "OLP\n(n = 20)", "PTP\n(n = 21)")) + 
    scale_y_continuous(limits = c(0, 35), expand = c(0, 0)) + 
    labs(title = "The Effect of Planning Time on the Amount of Syntactic Variety", 
         subtitle = "in three different conditions", 
         x = "", y = "Syntactic Variety\n") + 
    # change the overall theme (using one of the presets) and text size
    theme_bw(base_size = 14) + 
    # change theme elements 
    theme(panel.grid.major.x = element_blank(), # remove grid for x-axis
          axis.ticks.x = element_blank(), # remove x-axis ticks
          title = element_text(size = rel(.7))) # adjust the title size

Saving the file

ggsave(filename = "planningtime.png")
## Saving 7 x 5 in image
# or, provide more specifications
ggsave(filename = "planningtime.jpeg", width = 6, height = 4, units = "in", dpi = 800)
  • ggsave saves the last plot you created in the default size with 300 dpi.
  • filename: Include the file format extention to your file name. This is the only requirement for this function - everything else has default values. Certain image file format might not work with Mac OS (e.g., tiff).
  • The file will be saved in your working directory by default (unless you provide the path argument).

Extra: Violin plot alternative

ggplot(ell, aes(x = condition, y = variety, fill = condition)) + 
    geom_violin(scale = "count", width = .3, alpha = .7) + 
    geom_boxplot(width = .1, outlier.size = -1, fill = "white") + 
    geom_dotplot(binwidth = .5, binaxis = "y", stackdir = "center", fill = "black", alpha = .5) + 
    stat_summary(fun.y = "mean", geom = "point", shape = 23, size = 4, alpha = .7) + 
    labs(title = "Effect of Planning Time on the Amount of Syntactic Variety", 
         x = "", y = "Syntactic Variety\n") + 
    scale_x_discrete(labels = c("NP\n(n = 19)", "OLP\n(n = 20)", "PTP\n(n = 21)")) + 
    scale_y_continuous(limits = c(0, 35), expand = c(0, 0)) + 
    scale_fill_viridis_d() + 
    theme_bw(base_size = 14) + 
    theme(legend.position = "none", panel.grid.major.x = element_blank(), 
          axis.ticks.x = element_blank())

Example 2: Visualizing bivariate relationship: Correlation

STEP 1: Data (mainpulated)

str(obarow)
## 'data.frame':    67 obs. of  10 variables:
##  $ id       : int  3 2 6 7 1 10 5 8 11 14 ...
##  $ gender   : Factor w/ 2 levels "female","male": 2 2 2 2 2 1 2 2 1 1 ...
##  $ grade    : int  1 1 1 1 1 1 2 2 2 2 ...
##  $ treatment: Factor w/ 4 levels "NMNP","NMYP",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ pretest  : int  15 11 13 14 13 14 18 16 15 17 ...
##  $ posttest : int  14 11 13 15 12 14 16 14 13 16 ...
##  $ gain1    : int  -1 0 0 1 -1 0 -2 -2 -2 -1 ...
##  $ gain2    : int  -2 1 0 1 0 5 4 2 1 -1 ...
##  $ gain3    : int  0 0 1 3 -1 -1 0 0 1 1 ...
##  $ gain4    : int  0 3 1 0 1 -1 0 0 0 1 ...

Let’s visualize the correlation between pretest and posttest

ggplot(data = obarow, mapping = aes(x = pretest, y = posttest))

STEP 2: Elements

Visualize data points:

ggplot(obarow, aes(x = pretest, y = posttest)) + 
    geom_point()

Add a line that fits the data using stat_smooth function:

ggplot(obarow, aes(x = pretest, y = posttest)) + 
    geom_point() + 
    stat_smooth() # default is loess line
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'

Specify which method to use:

ggplot(obarow, aes(x = pretest, y = posttest)) + 
    geom_point() + 
    stat_smooth(method = "lm")

STEP 3: Attributes

Jitter the dots:

ggplot(obarow, aes(x = pretest, y = posttest)) + 
    geom_point(position = position_jitter()) + 
    stat_smooth(method = "lm")

Modify the trend line:

ggplot(obarow, aes(x = pretest, y = posttest)) + 
    geom_point(position = position_jitter()) + 
    stat_smooth(method = "lm", se = FALSE, lty = "dashed", color = "firebrick")

Let’s try a larger data set

Reading and writing data (fake data based on real data)

rw <- read.csv("readwrite2.csv")
str(rw)
## 'data.frame':    1997 obs. of  4 variables:
##  $ id     : int  3070 1306 83 2486 1938 397 977 1322 2414 156 ...
##  $ reading: int  18 20 14 16 14 20 20 16 21 21 ...
##  $ writing: int  21 15 14 20 15 18 9 14 15 24 ...
##  $ level  : int  3 3 2 4 2 4 2 1 3 5 ...

We have three variables: reading score = reading, writing score = wrting, group = level

Let’s see if there’s any correlation between reading and writing scores

ggplot(rw, aes(x = reading, y = writing)) + 
    geom_point(position = position_jitter())

ggplot(rw, aes(x = reading, y = writing)) + 
    geom_point(position = position_jitter()) + 
    stat_smooth(method = "lm")

Add proficiency level to the aesthetics:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter())

Change the level variable to factor and provide labels:

rw$level <- factor(rw$level, labels = c("Novice", "Low-IM", "High-IM", "Advanced", "Superior"))

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3)

Fitting trend lines:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3) + 
    stat_smooth(method = "lm", se = FALSE)

This will fit a line for each group because we have specified group (level)

To fit a line for overall data, but still keep the group colors, move the color aesthetic to lower level function

ggplot(rw, aes(x = reading, y = writing)) + 
    geom_point(aes(color = level), position = position_jitter(), alpha = .3) + 
    stat_smooth(method = "lm", se = FALSE)

Notice Simpson’s paradox here. When you fit an overall trend line, it looks like there is a positive correlation between reading and writing scores, but if you look at proficiency groups, there is no correlation or negative correlation. In this case, the trend seems that the scores increase in general as proficiency advances.

Color scale

Using viridis scheme:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3, size = 1) + 
    stat_smooth(method = "lm", se = FALSE) + 
    scale_color_viridis_d()

Using color brewer palettes:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3, size = 1) + 
    stat_smooth(method = "lm", se = FALSE) + 
    scale_color_brewer(palette = "Set1")

Axis scale

Change the x- and y-axes:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3, size = 2) + 
    stat_smooth(method = "lm", se = FALSE) + 
    scale_color_viridis_d() + 
    scale_x_continuous(limits = c(0, 30), expand = c(0, 0)) + 
    scale_y_continuous(limits = c(0, 30), expand = c(0, 0))

Use coordinate that applies 1:1 scale to be accurate with the visuals:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3, size = 2) + 
    stat_smooth(method = "lm", se = FALSE) + 
    scale_color_viridis_d() + 
    scale_x_continuous(limits = c(0, 30), expand = c(0, 0)) + 
    scale_y_continuous(limits = c(0, 30), expand = c(0, 0)) + 
    coord_equal()

This code is also possible:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), alpha = .3, size = 2) + 
    stat_smooth(method = "lm", se = FALSE) + 
    scale_color_viridis_d() + 
    # specify the x- and y-axis limits
    coord_equal(xlim = c(0, 30), ylim = c(0, 30), expand = FALSE)

Final, polished version:

ggplot(rw, aes(x = reading, y = writing, color = level)) + 
    geom_point(position = position_jitter(), 
               alpha = .3, size = 1) + 
    stat_smooth(method = "lm") + 
    stat_smooth(method = "lm", color = "black", se = FALSE) + 
    scale_color_viridis_d() + 
    coord_equal(xlim = c(0, 30), ylim = c(0, 30), expand = FALSE) + 
    labs(x = "\nReading score", y = "Writing score\n", 
         title = "Correlation between reading and writing scores", color = "Proficiency level") + 
    theme_bw()

save

ggsave("correlation.png", width = 6, height = 5, dpi = 600)

Extra

Facet: Create subplots

ggplot(rw, aes(x = reading, y = writing)) + 
    geom_point(position = position_jitter(), alpha = .1) + 
    stat_smooth(method = "lm") + 
    coord_equal() + 
    facet_wrap(. ~ level, nrow = 1) + 
    labs(title = "Relationship between reading and writings score by proficiency level", 
         x = "Reading score", y = "Writing score\n") + 
    theme_bw()

Example 3: Graphics for Repeated Measures

Creating parallel coordinate plot for pre- and post-tests

STEP 1: Data

str(obarow)
## 'data.frame':    67 obs. of  10 variables:
##  $ id       : int  3 2 6 7 1 10 5 8 11 14 ...
##  $ gender   : Factor w/ 2 levels "female","male": 2 2 2 2 2 1 2 2 1 1 ...
##  $ grade    : int  1 1 1 1 1 1 2 2 2 2 ...
##  $ treatment: Factor w/ 4 levels "NMNP","NMYP",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ pretest  : int  15 11 13 14 13 14 18 16 15 17 ...
##  $ posttest : int  14 11 13 15 12 14 16 14 13 16 ...
##  $ gain1    : int  -1 0 0 1 -1 0 -2 -2 -2 -1 ...
##  $ gain2    : int  -2 1 0 1 0 5 4 2 1 -1 ...
##  $ gain3    : int  0 0 1 3 -1 -1 0 0 1 1 ...
##  $ gain4    : int  0 3 1 0 1 -1 0 0 0 1 ...
  • Dependent/Outcome/Response variable: pretest and … posttest?
  • Independent/Predictor variable: treatment

The two different tests are repeated measures of one variable: test score

Reshape the data so that:

  • Dependent/Outcome/Response variable: score
  • Independent/Predictor variables: treatment & test

Transform the data

ob2 <- tidyr::gather(data = obarow, key = test, value = score, pretest, posttest)
  • key: the new variable with categories (i.e., the two types of tests)
  • value: the values associated with the new categorical variable (i.e., test score)
str(ob2)
## 'data.frame':    134 obs. of  10 variables:
##  $ id       : int  3 2 6 7 1 10 5 8 11 14 ...
##  $ gender   : Factor w/ 2 levels "female","male": 2 2 2 2 2 1 2 2 1 1 ...
##  $ grade    : int  1 1 1 1 1 1 2 2 2 2 ...
##  $ treatment: Factor w/ 4 levels "NMNP","NMYP",..: 1 1 1 1 1 1 1 1 1 1 ...
##  $ gain1    : int  -1 0 0 1 -1 0 -2 -2 -2 -1 ...
##  $ gain2    : int  -2 1 0 1 0 5 4 2 1 -1 ...
##  $ gain3    : int  0 0 1 3 -1 -1 0 0 1 1 ...
##  $ gain4    : int  0 3 1 0 1 -1 0 0 0 1 ...
##  $ test     : chr  "pretest" "pretest" "pretest" "pretest" ...
##  $ score    : int  15 11 13 14 13 14 18 16 15 17 ...

What are we mapping?

ggplot(data = ob2, mapping = aes(x = test, y = score))

Is there anything that appears to be wrong with the coordinates? Yes, because the levels of test is arranged alphabetically, posttest appears before pretest.

Specify the order the factor levels:

ob2$test <- factor(ob2$test, levels = c("pretest", "posttest"))
ggplot(data = ob2, mapping = aes(x = test, y = score))

Now it looks right. Same for any discrete levels of a given variable (treatment groups, etc.)

STEP 2: Elements: which geometric layers need to be visualized?

  1. We want a line graph that shows score change from pretest to posttest
ggplot(ob2, aes(x = test, y = score)) + 
    geom_line()

What went wrong?

  • geom_line() is just connecting data points within each test category.
  • We want one line for each individual, one that indicates the change from pre- to post-test.

Group the scores for each individual by assigning the id variable to group. Either of these works:

ggplot(ob2, aes(x = test, y = score, group = id)) + 
    geom_line()

ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id))

This means, you need a variable called ID for creating this type of graphs. If you don’t have one create it in the original dataset before transforming the data into a long form.

  1. We want to overlay group means

Just as we provided the id variable to group argument, we need to group the data by the treatment groups to plot the group means.

ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id)) + 
    stat_summary(aes(group = treatment), fun.y = "mean", geom = "line", size = 2)

  1. We could also add CIs
ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id)) + 
    stat_summary(aes(group = treatment), fun.y = "mean", geom = "line", size = 2) + 
    stat_summary(aes(group = treatment), fun.data = "mean_cl_normal", 
                 geom = "pointrange", size = 1)

STEP 3: Attributes: refine the visuals

  1. Add colors

by treatment group:

ggplot(ob2, aes(x = test, y = score, color = treatment)) + 
    geom_line(aes(group = id)) + 
    stat_summary(aes(group = treatment), fun.y = "mean", geom = "line", size = 2)

Not necessary because it is hard to decode each line.

just for the mean:

ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id)) + 
    stat_summary(aes(group = treatment, color = treatment), 
                 fun.y = "mean", geom = "line", size = 2)

  1. Add CIs and adjust positions to avoid overplotting:
ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id), color = "grey70") + 
    stat_summary(aes(group = treatment, color = treatment), fun.y = "mean", 
                 geom = "line", size = 2, position = position_dodge(.1)) + 
    stat_summary(aes(group = treatment), fun.data = "mean_cl_normal", 
                 geom = "pointrange", size = 1, position = position_dodge(.1)) + 
    scale_color_brewer(palette = "Dark2")

STEPS 4&5: Polish it up:

ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id), color = "grey70") + 
    stat_summary(aes(group = treatment, color = treatment), fun.y = "mean", 
                 geom = "line", size = 2, position = position_dodge(.1)) + 
    stat_summary(aes(group = treatment), fun.data = "mean_cl_normal", 
                 geom = "pointrange", size = 1, position = position_dodge(.1)) + 
    scale_color_brewer(palette = "Dark2") + 
    scale_x_discrete(expand = c(.2, .2), labels = c("Pre-test", "Post-test")) + 
    scale_y_continuous(limits = c(0, 30)) + 
    labs(x = "", y = "Vocabulary Score\n", title = "Score Changes after Treatment") + 
    theme_bw() + 
    theme(legend.position = "none", panel.grid.major.x = element_blank(), 
          axis.ticks.x = element_blank())

Alternative: Faceted graphics

ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(aes(group = id), color = "grey70") + 
    stat_summary(aes(group = treatment), fun.y = "mean", geom = "line", size = 2) + 
    stat_summary(aes(group = treatment), fun.data = "mean_cl_normal", 
                 geom = "pointrange", size = 1) + 
    scale_x_discrete(expand = c(.2, .2), labels = c("Pre-test", "Post-test")) + 
    scale_y_continuous(limits = c(0, 30)) + 
    labs(x = "", y = "Vocabulary Score\n", title = "Score Changes after Treatment") + 
    facet_grid(. ~ treatment) + 
    theme_bw() + 
    theme(panel.grid.major.x = element_blank())

Extra:

Use differen color lines for individuals who had positive gain scores after treatment.

ggplot(ob2, aes(x = test, y = score)) + 
    geom_line(data = dplyr::filter(ob2, gain1 > 0), aes(group = id), color = "forestgreen") + 
    geom_line(data = dplyr::filter(ob2, gain1 <= 0), aes(group = id), color = "grey70") + 
    stat_summary(aes(group = treatment), fun.y = "mean", geom = "line", size = 2) + 
    stat_summary(aes(group = treatment), fun.data = "mean_cl_normal", 
                 geom = "pointrange", size = 1) + 
    scale_y_continuous(limits = c(0, 30)) + 
    labs(x = "", y = "Vocabulary Score\n", title = "Score Changes after Treatment") + 
    facet_grid(. ~ treatment) + 
    theme_bw() + 
    theme(panel.grid.major.x = element_blank())