IST 387: Intro to Applied Data Science | Jasmina Tacheva

Practice Midterm Exam

Instructions: Find a quiet place to work where you will not be interrupted for at least 60 minutes.
Just as your lab and homework assignments, your exam will be graded based on the R script you
submit, so make sure the R script you create for this practice exam is as detailed as possible, with
clear comments accompanying each new block of code. It is important that you know what each line
of code you use is doing and what its syntax is.
Once you are done with all questions, save your R script. I will go over this practice exam during
next week’s office hours (6:30-8:30 pm on October 19th):
https://syracuseuniversity.zoom.us/j/92651366973
If you cannot join the Zoom meeting but would like to ask a question about the practice test, please
email me.

Dataset Description: The dataset used in this practice exam is called testData.csv and can be
found on Blackboard. Download it to a folder on your computer and then set your R working
directory to that folder the way we did in class and your lab section last week:
setwd("C:\\Path\\to\\Folder") # Change to the folder containing your twitterData.csv file
Read the file into a dataframe called “data” using the read_csv() function from the tidyverse
package.
Inspect your dataframe using the appropriate R function – it contains the following variables:
Twitter user ID, number of followers a user has, number of users the focal user follows, total
number of tweets a user has posted, and the user’s state of residence. Each observation, aka row, in
the dataframe therefore represents the record of a unique Twitter user.
Don’t forget to “library” the appropriate R packages you think you will need to complete the tasks
below.

Research Questions (aka, your tasks):
1. Describe the number of followers variable using descriptive statistics provided by R. Do the
same for the number of users followed variable.
Hint: What function(s) have we used so far to summarize variables in a dataset?
2. Describe the shape of the distribution for number of followers. Do the same for number of
users followed.
Hint: How do we represent a distribution in R? Perhaps a histogram might help? Are the
variables normally distributed? Or are their distributions right- or left-skewed?
3. On average, do the focal Twitter users in this dataset follow more accounts, or are they
followed by more accounts?
Hint: Think of a statistical measure for each of the two variables that can help you make this
determination. Perhaps looking at a measure of central tendency would help?
4. Create a new variable that represents the difference in number of followers and number of
users followed for each focal Twitter user, aka observation, aka row. Describe the shape of
the distribution of this new variable.
5. Create a scatterplot of the number of followers and number of users followed. Clearly label
your axes so they are more descriptive. Does the scatterplot show any pattern or
relationship?
6. Generate a linear model to predict the number of followers based on the number of tweets
and the number of users followed. Generate another linear model to predict the number of
users followed based on the number of tweets and the number of followers a user has.
7. Interpret the coefficients of the statistically significant predictors in the two models.
Comment on the quality of each model. Which model is better – explain.
IST 387: Intro to Applied Data Science | Jasmina Tacheva

8. If you come across a Twitter user with 541 followers and 1128 tweets, what would be your
model’s prediction about the number of people this user follows?
9. What would be your best guess about the number of people a user with 0 followers and 0
tweets follows?
10. Generate a map of the average number of tweets in each state where each state is shaded
depending on its average number of tweets.
Hint#1: You may need to use the aggregate() function from your ggmap HW, since the
current level of analysis in your data is individual users and you now want the level of
analysis to be states.
Hint#2: The FUN (aka “function”) part of the aggregate() code in your HW may need to be
modified – in your HW, we wanted to add up the income of individual ZIP code areas when
aggregating, that’s why we used “sum,” but now we want the average – what might you
want to replace “sum” with in that case?
11. Your map likely looks weird - what argument can you add to your map code to make sure
states with no data still appear on the map, with an outline color and a fill color?
12. Based on your analysis so far, do you think the number of followers and the number of
users followed are related?

欢迎咨询51作业君