Description
Read Graphical Data Analysis with R, Ch. 3
Grading is based both on your graphs and verbal explanations. Follow all best practices as discussed in class.
The datasets in this assignment are from the ucidata package which can be installed from GitHub. You will
first need to install the devtools package if you don’t have it:
install.packages(“devtools”)
then,
devtools::install_github(“coatless/ucidata”)
1. Abalone
[18 points]
Choose one of the numeric variables in the abalone dataset.
a) Plot a histogram of the variable.
b) Plot histograms, faceted by sex, for the same variable.
c) Plot multiple boxplots, grouped by sex for the same variable. The boxplots should be ordered by
decreasing median from left to right.
d) Plot overlapping density curves of the same variable, one curve per factor level of sex, on a single set
of axes. Each curve should be a different color.
e) Summarize the results of b), c) and d): what unique information, specific to this variable, is provided
by each of the three graphical forms?
f) Look at photos of an abalone. Do the measurements in the dataset seem right? What’s the issue?
2. Hepatitis
[6 points]
a) Draw two histograms of the age variable in the hepatitis dataset in the ucidata package, with
binwidths of 5 years and boundary = 0, one right open and one right closed. How do they compare?
b) Redraw the histogram using the parameters that you consider most appropriate for the data. Explain
why you chose the parameters that you chose.
3. Glass
[18 points]
a) Use tidyr::gather() to convert the numeric columns in the glass dataset in the ucidata package to
two columns: variable and value. The first few rows should be:
variable value
1 RI 1.52101
2 RI 1.51761
3 RI 1.51618
4 RI 1.51766
1
Use this form to plot histograms of all of the variables in one plot by faceting on variable. What patterns
do you observe?
For the remaining parts we will consider different methods to test for normality.
b) Choose one of the variables with a unimodal shape histogram and draw a true normal curve on top on
the histogram. How do the two compare?
c) Perform the Shapiro-Wilk test for normality of the variable using the shapiro.test() function. What
do you conclude?
d) Draw a quantile-quantile (QQ) plot of the variable. Does it appear to be normally distributed?
e) Use the nullabor package to create a lineup of histograms in which one panel is the real data and the
others are fake data generated from a null hypothesis of normality. Can you pick out the real data? If
so, how does the shape of its histogram differ from the others?
f) Show the lineup to someone else, not in our class (anyone, no background knowledge required). Ask
them which plot looks the most different from the others. Did they choose the real data?
g) Briefly summarize your investigations. Did all of the methods produce the same result?
4. Forest Fires
[8 points]
Using the forest_fires dataset in the ucidata package, analyze the burned area of the forest by month.
Use whatever graphical forms you deem most appropriate. Describe important trends.
