Interesting Comments from Fall 2019 for Gov 1005

  • Find ways to motivate us to do the readings before class.

  • I think one of the things that would be helpful would be to start learning Shiny earlier.

  • I think we could implement pair programming for problem sets or other assignments, especially since it is so widely used in industry.

  • Include some poetry in this course

  • Staff needs to know Shiny

  • I felt we would be working on the same problem for like 4 classes in a row, restarting at the beginning each time. Maybe having a forked repo with a clean version of what we covered in the last class, reviewing that briefly and then moving on (more like what we are doing with machine learning) could be a good idea.

  • I think having a coherent workflow throughout the problem sets would be helpful (e.g. A. Download data / B. Clean Data / C. Prep / D. Analysis / E. Writing out solutions…)

  • The difference between what we do during class and what we do on our own is too severe. The course would benefit from increasing the difficulty of the work we do in class. Often, I feel the in-class instructional time is too slow-paced or too redundant in the material covered. Students would probably fare better on problem sets and exams if the in-class work challenged them a bit more.

  • One non-class material thing I think would be beneficial to change is to switch the seating around a little more frequently in the beginning of the class. We stayed in the house sections for the first 2 months, which was fine, but also made it harder to “meet everyone in the class” like one of the goals is. I think maybe switching it up every couple of weeks would better facilitate that goal.

  • I actually think you don’t need to cut as much of the datacamp as you’re cutting. I think the datacamp only feels like too much because we don’t actually use the stuff we learn in it that often, and then it feels like why are we spending so much time on something not really relevant to the in-class work we’re doing. If problem sets were more reflective of the datacamp (by coordinating their content a little more) then it would at least feel more useful.

  • I would not take out those data camp assignments; for datacamp, I would honestly keep all the assignments you currently give, but don’t assign entire chapters due all at one week. Assigning mini sections every two days would honestly make it much easier for students to plan out timewise how much datacamp they should be doing, rather than doing all of the chapters (4 hours worth) the night they are due.

  • I would like to argue for keeping the DataCamp assignment on text analysis because my final project is relying on that one heavily. Consider maybe assigning the first chapter of each of these or just a portion so that students can see if they want to use anything in these places on their final projects.

  • If you are taking out those data camps, please make the shiny datacamp required next year! It made a world of difference in my ability to complete my final project and I don’t think enough people watched it.

  • But what would be sick– you walk us through 3 small but crazy projects that include Machine Learning. A different way to give us a taste of what’s cool. One project is for campaigns, one is for sports teams, one is for Harvard. You write out a lot of the data already, and it’s up to us to finish things, especially the code that we already understand. It gives us an opportunity to use our skills on a new and cool topic, and gives us a taste so that we can tell everyone, “we know machine learning!” But it doesn’t assume we can learn this skill in 3 days.

  • I do think it (machine learning) is a bridge too far for an intro course. Perhaps it would be helpful to have different tracks for people who are beginners in R versus feel more comfortable/have more experience. The more advanced track could include more machine learning.

David Kane
Preceptor in Statistical Methods and Mathematics
comments powered by Disqus