Teaching at Monsanto

Will Trimble writes:

In the Python room, we had about 18 learners the first day, slightly fewer the second day, and between 5-10 participating by Webex. This was the most advanced audience I've seen; they were bored by the two-letter shell commands, and could have probably gone through find, for, and grep in a quarter day.

We covered the shell with the standard lessons, and Python with the inflammation dataset; on day 2 we did Git in pairs, and since they quickly finished resolving conflicts locally and pushing to their neighbor's repository we did a central repository fork-and-pull request (11 forks, 7 pull requests) in the remaining time.

On day 2 we worked through building a simple Python command line script, and checked it in every time it changed. The end of the second day we went through the first of the Python-mosquitoes (Pandas) notes and the matplotlib+basemaps notebook by Nikolay Koldunov. We didn't get good coverage of either testing or data structures (dicts in particular, which might be unfamiliar to R and Fortran users) and did not attempt SQL.

When polled, the audience was glad for the instruction, though there were some technical issues:

There were widespread reports that neither more nor less was working for some reason under Git Bash, and cat has been written out of the lesson for simplicity reasons.
To run the fancy graphics notebook we tried to run conda, and hit a snag where Git Bash + conda did not work for most users without first disabling SSH for conda. To do this, make a file called ".condarc" in C:\Users\(your-user-id) containing:
```
# Disable ssl verification. The default is True.
ssl_verify: False
```
then conda install basemaps should work.

Asela Wijeratne writes:

From the feedback we got, I believe that overall, we did well. I think the Git, SQL and shell lessons went fairly well and Xu did an excellent job teaching SQL with a very short notice. For the R lessons, we covered some of the basics (like data types, data structures and subsetting etc), creating function in R, and the intro to ggplot.

However, for R lessons, we could have done better (partly, because this is my first time teaching it). In addition, it appeared (from the discussion we had with the participants and also from the pre-workshop survey), most participants were well versed in R. So I expected that I could just introduced some of basics and then moved to more advance topics like creating functions and control of flow. Once I started to teach, I realized that most people needed a refresher on R basics and eventually, I didn't have time to do all the topics and examples that I was planning to cover. In the end, R lessons should have been two sessions.

In addition, as Xu pointed out, perhaps there should be a different way to assess the knowledge of the participant (for e.g., instead of asking whether they know R, we could ask them to complete a task without external resources).

The second problem, which might be specific to companies and government institutes, is that there are certain restrictions where we can post data; for example, I posted the data on Google drive and I was told that the participants could not access it. In addition, we should explicitly mention that which version of R and RStudio we are planning to use (this may apply for others software as well). Some attendees were using older versions of R and it was not easy to switch as they needed to get permission to do so.

Xu Fei writes:

The R session was overall challenging but smooth. Our pre-workshop questions didn't capture the background level of participants (more specifically on the Shell and R) very well this time. I think it made teaching such a diverse audience challenging but I think Asela managed it pretty well. Also, like Asela mentioned, the installation note for R may need an update, as I've seen enough cases where installation of a specific library of R failed due to version issues.

The Git lesson did not go as smoothly as in the other room. The good thing was that people were asking a lot of good questions (and from WebEx). But I got caught in the headless state (consequence of being too creative?) and created a bit of confusion when I was trying to resolve a conflict. We did not have time for all the participants to resolve a conflict, and we certainly did not cover pull requests.

SQL was pretty standard and I had time to cover all the topics, even with an additional session on how to use SQLite in R. I think it was a good way to end the whole workshop as it made people connect with what they just learned in the morning. Again, our audience didn't have much experience with Git and SQL and it made it a bit easier to control the flow.