Reflection on Independent Projects
I was the faculty supervisor for two independent projects. I include the syllabus for both projects, one in Spring 2016 and one in Spring 2017. In the Fall prior to each Spring, the students, for which these projects were created, were enrolled in my Introduction to Statistics course. Simon's Rock doesn't have a statistics program. Nor does it have a data science program. For students who enjoy Intro to Stats but are not inclined to take more calculus, there is no where for them to go next, and so many of them do not continue in any quantitative course. When a student approaches me about what they can do next, I always suggest an independent project. Not all students are inclined to take on more work. I have only three students who successfully took on the extra work. I discuss two of the projects.
In the Spring of 2016, I supervised an independent project for the first time. The student I advised had taken Introduction to Statistics with me the semester prior. They enjoyed the class very much (both the statistics and the R programming) and wanted to continue doing more data analysis. This student was a Psychology senior. I agreed to an independent project in data science because this is intended to be a student-led course, and I believed a senior in college would have a work ethic to make a successful project. I wanted to
keep the student on task to learning more data science through homework assignments via the learning platform datacamp
quiz the student in order to determine whether they learned from the homework
have the student create a project which utilized their skill from datacamp as well as the skill from the statistics course.
The syllabus and the goals above were very ambitious compared to other independent projects I have seen. We did follow the syllabus very closely. However, there were some issues with the goals I set forth.
The student did do R homework assignments I selected through datacamp. Because datacamp has created coding assignments that are very guided, i.e. there are videos that teach and hints provided, I believed this would be an independent activity and I would have to do very little. This turned out to not be true. The student did not watch the videos, and the hints are often not very helpful. In fact, the system, at that time, was designed so that a student could be given the answer for deducted points, but then the next day go in and resubmit for full points. In the end, I did not count this homework for much toward the final grade, believing the exposure was beneficial even though the student didn't learn to think critically when programming in R.
The student did well on the quizzes, indicating they did learn from the flawed system.
The project was the most difficult for the student.
What the student needed to do was not covered in the homework assignments, despite my best efforts to predict what they might need. For example, the data set chosen was online and not in a clean format. The student needed to know how to scrap data and clean it.
For the particular topic chosen, very little intro to statistics material was needed for the final analysis. Although these could certainly be used, there was no need for p-values or confidence intervals beyond what the student already used.
In the end, however, the project turned out to be relatively successful. Clearly, the formatting of this report could be improved as well as the number of outside graphs and charts could have been limited to one. The project seems rushed. However, the student did quite a bit of outside research on other analyses for this data. That is, the student did not limit their report to only what they found in the data set themselves. In addition, the student did their best to apply prior statistics knowledge, using the Chi-Square Test to determine whether a relationship existed between race of the executed and the year the individual is executed. The student investigated the data set with a series of questions they wanted to ask and was able to answer many of them.
In the Spring of 2017, I supervised an independent project for the second time. The student I advised had, like the first student, taken Introduction to Statistics with me the semester prior. This student was not a senior, but a sophomore. They planned to continue taking mathematics and go into statistics. I agreed to an independent project in data science because this is the student's passion and they would have the dedication to make a successful project. Based on my experience from the first time, I changed my goals for the independent project.
I would still keep the student on task, but I would not do this with problems designed by datacamp. Rather, the student would begin the project right away and homework assignments would be the research needed to understand how to complete the steps needed for the final project.
Quizzes are completed verbally by questions I ask the student during our weekly sessions.
The final project report will need to be formatted more professionally, so I required the student to learn and to use the package known as R Markdown.
I want the student to demonstrate that they have learned and are an expert on their research by presenting their findings and answering an audience's questions. The student did so with relative success.
In the end, this project was also successful in ways the first was not. Clearly, the formatting has improved. In addition, the student learned to use several new R functions in order to obtain meaningful graphs. However, this report lacks where the first did do well. First, no statistical inference method is used. Second, more outside research could improve the conclusion section. Though, this project is equivalent to many exploratory data analysis projects viewed online.
I believe the failures of the projects come from the lack of experience with scientific research and writing. Both students learned from the projects, and this is significant. In the future, I may provide examples of good data science projects so that the students can model them.