Statistics and Capstone Project

21 May 2018

This was the last session of the series. The plan with this session was to make this a two parter:

  • Part 1 (~30 minutes): a brief and whirlwind tour of linear modelling using lm() as a way to introduce R’s syntax for formulae and hint at the usefulness of R’s base stats package

  • Part 2 (~1.5 hours): a Capstone project using where learners were given some raw data and told to go wild analyzing it using the skills we’ve covered in the series so far. The dataset was a fun one: one of the students, Yannay, has tracked two years of his finances down to individual purchases.

It ended up that we held a co-working session a week after the formal session which some people came to so as to keep working on the Capstone project, so learners got an additional 2 hours to work on the project.

Overall, I have mixed feelings about this session: I tried to do too much in one sitting, and the learners probably weren’t far enough along to jump into a project like this on their own, but on the whole people seemed to enjoy the chance to play with an interesting dataset and practice their skills.

My take-aways:

  • I didn’t teach statistics, I just taught how to run linear models in R, some of the learners were completely lost in Part 1. If I were to do this again, I’d have an longer session devoted to linear modelling… maybe even two.
  • Because I tried crammed linear modelling into 30 minutes, it was entirely a show-and-tell style session. This meant that even if people had enough of a background in statistics to follow along, they didn’t get any time to practice. I was hoping those people who were keen to practice their stats would use the Capstone project time to do that, but I think they spent most of the time working on the data wrangling and visualization tasks.
  • As much as I wanted learners to dive into the data, and go back to earlier lessons to recall various clean up and data wrangling and visualization techniques to apply them, most weren’t able to. It was just too much of a stretch to, say, ask of learners to remember the concept and syntax to filter() data from two lessons back. Worse, when it came to some of the more abstract concepts like group_by and summarise – which I think people started to understand during the lesson – learners were totally out of their depth without some nudging.
  • As a result, in the first session, I ended up walking people through how to plot the amount of money spent each month. I think learners found that useful and used that code to answer other related questions (mean spending per month, spending per year, spending across categories, etc.). Note to self: always give people starter code.
  • Dates are hard. I hadn’t taught anything about date manipulation, and the dataset we used had numeric Year, Month, and Day columns. I ended up just gesturing at the magic of the ISOdate() function and letting learners know about the lubridate package if they wanted to dive into date handling more fully.
  • The co-working session was great. Only a few people attended, but I think they made good use of the extra time to play around with the dataset, and we had some good interactive discussions about dplyr, and ggplot. I think if I ran these types of co-working sessions once a month, along with lessons, everyone would have practiced a whole lot more.