Table 1
We retrieved our data from Table 1 on page 15 in the following article by Sullivan and Artiles:
Sullivan, A.L., & Artiles, A.J. (2011). Theorizing racial inequity in special education: Applying structural inequity theory to disproportionality. Urban Education, 46, 1526-1552.
These data were entered into a Microsoft Excel file in “long” format.
## Warning: cannot open file
## '/Users/aleksiskincaid/Downloads/RRRLEADataFrame.csv': No such file or
## directory
## Error: cannot open the connection
## Disability StateRRR Ethnicity LEALow LEAHigh
## 1 SPED 1.22 AA 17.50 23.8
## 2 SLD 1.38 AA 27.50 29.7
## 3 SLI 0.82 AA 43.75 9.2
## 4 ED 1.48 AA 37.50 18.9
## 5 MIMR 2.42 AA 35.00 27.0
## 6 OHI 0.85 AA 48.75 10.8
## 7 Autism 0.89 AA 43.13 6.6
## 8 HI 0.76 AA 35.00 1.1
## 9 MD 1.30 AA 37.50 9.7
## 10 MOMR 1.53 AA 41.25 10.8
## 11 OI 0.71 AA 35.63 4.9
## 12 TBI 1.00 AA 31.88 1.6
## 13 VI 0.86 AA 27.50 2.7
## 14 SPED 0.96 HI 11.18 9.7
## 15 SLD 1.19 HI 9.41 19.5
## 16 SLI 0.86 HI 21.76 9.7
## 17 ED 0.37 HI 57.65 5.9
## 18 MIMR 1.38 HI 21.18 15.7
## 19 OHI 0.40 HI 50.59 3.2
## 20 Autism 0.42 HI 46.47 3.2
## 21 HI 1.12 HI 20.00 8.6
## 22 MD 0.95 HI 25.29 5.9
## 23 MOMR 1.35 HI 24.71 15.1
## 24 OI 0.71 HI 27.65 4.3
## 25 TBI 0.75 HI 20.59 5.4
## 26 VI 0.71 HI 23.53 2.7
## 27 SPED 1.26 NAM 16.46 23.8
## 28 SLD 1.71 NAM 21.34 39.5
## 29 SLI 0.90 NAM 39.02 15.7
## 30 ED 0.65 NAM 51.22 11.4
## 31 MIMR 1.49 NAM 20.73 16.8
## 32 OHI 0.45 NAM 59.15 5.4
## 33 Autism 0.03 NAM 39.63 3.8
## 34 HI 1.71 NAM 29.88 4.3
## 35 MD 1.70 NAM 21.34 11.4
## 36 MOMR 1.41 NAM 43.29 10.8
## 37 OI 1.00 NAM 37.20 4.3
## 38 TBI 1.50 NAM 28.66 4.9
## 39 VI 1.57 NAM 28.66 3.8
## 40 SPED 0.58 API 53.10 9.7
## 41 SLD 0.42 API 68.28 5.9
## 42 SLI 0.79 API 53.79 9.7
## 43 ED 0.21 API 79.31 2.7
## 44 MIMR 0.69 API 61.38 4.9
## 45 OHI 0.40 API 73.10 3.2
## 46 Autism 0.11 API 53.10 5.9
## 47 HI 1.71 API 35.17 5.9
## 48 MD 0.07 API 52.41 5.4
## 49 MOMR 0.94 API 53.79 4.9
## 50 OI 1.14 API 38.62 5.9
## 51 TBI NA API 35.86 2.2
## 52 VI 1.00 API 33.79 2.7
The data were loaded into R with the following read.csv command:
Data=read.csv(file.path)
#file.path is the actual file path name on your computer.
#This will also name your dataframe "Data"
Next ggplot was loaded using:
library(ggplot2)
# If you don't have ggplot2 installed in R, simply download the package
# from the CRAN Repository
Now that we have the dataframe in R and ggplot2 is loaded, we will be able to graphically explore the data from Table 1. Sullivan and Artiles (2011) were interested in “investigating differential risk of educational disability across racial groups” (pp. 1526). Our goal is to graphically display what Sullivan and Artiles hoped their table would elucidate - differential risk of being placed in special education by race.
We decided to examine disability by state relative risk ratio based on ethnicity. The code is as follows.
ggplot(data=Data, aes(x=Disability, y=StateRRR, fill=Ethnicity))+
#We set x = Disability, Y = StateRRR and the fill for the points conditional on ethnicity.
#Remember to add + signs at the end of each line with ggplot code to add another layer
geom_point()
#geom_point() adds points to the graph for a scatterplot
At this point the graph looks like this:
## Warning: Removed 1 rows containing missing values (geom_point).
We thought the graph would be more interpretable if the points were different shapes and colors based on ethnicity. Also, as a matter of aesthetics, we wanted to eliminate the warning message that appeared due to one instance of missing data. The code now looks like this:
ggplot(data=Data, aes(x=Disability, y=StateRRR, fill=Ethnicity))+
geom_point(size=5, na.rm=TRUE, aes(color=Ethnicity, shape=Ethnicity))
#Within geom_point you can change the size with size=
#na.rm=TRUE eliminates the pesky error message
#aes is a handy function for aesthetics
#Here, we used aes() to change the color and shape conditional on ethnicity
The graph now looks like this:
At this point, we got picky about what shapes we wanted for ethnicity. In the above graph, the '+' symbol for Native American students is difficult to see on the graph. We also wanted to include lines for the cut scores that represent significant under- and over-representation of students compared to white students (centered at 1.0 on state relative risk ratio; StateRRR).
Specific shape's corresponding R-code numbers can be found on the R cookbook wesbite
We added a few lines of code to accomplish this:
ggplot(data=Data, aes(x=Disability, y=StateRRR, fill=Ethnicity))+
geom_point(size=5, na.rm=TRUE, aes(color=Ethnicity, shape=Ethnicity))+
scale_shape_manual(values=c(15,16,17,18))+
geom_hline(size=1,aes(yintercept=1.0))+
geom_hline(size=1,aes(yintercept=0.5, alpha=.7))+
geom_hline(size=1,aes(yintercept=1.5, alpha=.7))
#scale_shape_manual allows you to manually select the point shapes displayed on the R cookbook website
#geom_hline adds a line to the graph at the point that you determine
#again, the size function can be used
#alpha allows one to make the line transparent (0) to fully opaque (1)
#Lines were added at .5 and 1.5 to display what was considered disproportionate over- or under-representation of minority groups (anything between .5 and 1.5 is not considered disproportionate)
Now the graph looks like this:
We decided the graph would look better with lines connecting each individual ethinic group across disability categories to display trends. We also wanted to title the axes and the graph, as well as remove the grey background.
ggplot(data=Data, aes(x=Disability, y=StateRRR, fill=Ethnicity))+
geom_point(size=5, na.rm=TRUE, aes(color=Ethnicity, shape=Ethnicity))+
scale_shape_manual(values=c(15,16,17,18))+
geom_hline(size=1,aes(yintercept=1.0))+
geom_hline(size=1,aes(yintercept=0.5, alpha=.7))+
geom_hline(size=1,aes(yintercept=1.5, alpha=.7))+
#new code starts below
geom_line(size=1, aes(group=Ethnicity, color=Ethnicity))+
ylab("State Relative Risk Ratio")+
xlab("Disability Categories")+
ggtitle("State Level Relative Risk Ratio by Ethnicity")+
theme_bw()+
#added to remove the grey background
theme(axis.title.x=element_text(size=16)) +
theme(axis.title.y=element_text(size=16)) +
theme(plot.title=element_text(size=20))
#geom_line added a line to the graph, the aes function was used to separate the lines by ethnicity (group) and by color.
#labeling the graph follows pretty intuitive functions such as xlab for the x-axis label, ylab for the y-axis, and ggtitle to give the plot a title
#The theme function was used to change the size of the x and y labels as well as the the ggplot title
The graph now looks like this:
The graph looked like it was coming together and trends were becoming more interpretable than in the table of data. However, we wanted to clean the graph up and bit and make it look less chaotic. Thus, we created a white background, removed the minor gridlines, made the tickmark text larger, and changed the color of the major gridlines.
ggplot(data=Data, aes(x=Disability, y=StateRRR, fill=Ethnicity))+
geom_point(size=5, na.rm=TRUE, aes(color=Ethnicity, shape=Ethnicity))+
scale_shape_manual(values=c(15,16,17,18))+
geom_hline(size=1,aes(yintercept=1.0))+
geom_hline(size=1,aes(yintercept=0.5, alpha=.7))+
geom_hline(size=1,aes(yintercept=1.5, alpha=.7))+
geom_line(size=1, aes(group=Ethnicity, color=Ethnicity))+
ylab("State Relative Risk Ratio")+
xlab("Disability Categories")+
ggtitle("State Level Relative Risk Ratio by Ethnicity")+
theme_bw()+
theme(axis.title.x=element_text(size=16)) +
theme(axis.title.y=element_text(size=16)) +
theme(plot.title=element_text(size=20)) +
#new code starts below
theme(panel.grid.minor=element_blank())+
#to remove the minor gridlines
theme(panel.grid.major=element_line(color="grey78"))+
#to make the major gridlines grey
theme(axis.text=element_text(size=10))
#to make the tickmark labels larger
Voilà , the final graph:
This graph demonstrates the relative risk of being given a specific disability label for students in different ethnic groups compared to white students. The researchers also presented data about the percentage of local education agencies (LEAs) under (low) and over-representing (high) students in different ethnic groups across disability category compared to white students. The script for those graphs are similar to the previous one and will thus be explained in less detail.
ggplot(data=Data, aes(x=Disability, y=LEALow, fill=Ethnicity))+
#y changed to LEALow
ylab("Percentage of LEAs Underrepresenting Students")+
xlab("Disability Categories")+
ggtitle("Underrepresentation in Disability Category by Ethnicity")+
theme_bw()+
theme(axis.title.x=element_text(size=16)) +
theme(axis.title.y=element_text(size=16)) +
theme(plot.title=element_text(size=18))+
theme(panel.grid.minor=element_blank())+
theme(axis.text=element_text(size=10))+
theme(panel.grid.major=element_line(color="grey78"))+
geom_point(size=5, na.rm=TRUE, aes(color=Ethnicity, shape=Ethnicity))+
scale_shape_manual(values=c(15,16,17,18))+
geom_line(size=2, aes(group=Ethnicity, color=Ethnicity))
#hlines were omitted as there is no longer a singificant cutoff
This script produces this graph:
It is easy to graphically see trends in underrepresentation based on ethnic group as compared to white students. While the tabular data (seen in Table 1) were hard to compare and interpret, clear trends emerge in the graph.
The next graph has the same script as the previous except y is equal to LEAHigh rather than low. Thus, the script will not be reproduced.
Again, in graphical form, it is very easy to see emerging trends of which ethnic groups are over-represented in specific disability categories as compared to white students.