warrenagin
commented
6 years ago I had so many problems doing Anaconda installs on my computer (windows) that I can't imagine working through it in a class environment. Most of my class was on Apple.
I'd go talk to your school's engineering or comp sci department to see how they deal with the issue of setting up environments. Maybe the B-school (since they have to work with non-technical students) They might be of some help or have a stock solution you can use.
I used excel and on-line tools (Wolfram, online calculators) in part because I was too new to python and R to be comfortable teaching them, but also because I was concerned about the learning curve for students who have never coded before. I hear Smartsheets is a useful product too. You can pick up the skills you would need in about one or two hours. Sort & Filter, Pivot Table, Solver, and the Data Analysis plug-in. Plus knowing that you can drag cell entries across ranges.
paultopia
TrevorAustin
My current plan is to teach this course using Python, and to require students to install the Anaconda distribution (using the most recent version of Python 3 at the time), with the idea that most of the work will be done in Jupyter notebooks.
However, there are a few potential problems with this plan:
The perennial problem with library glitches, version incompatibilities, etc. The wonderful Vicki Boykis just wrote a wonderful blog post about all this mess. This can be a particular nightmare with the data stack, and it could waste a ton of time debugging and turn students off. One solution could be heavy use of virtual environments, but working with virtualenvs in jupyter notebook is a bit awkward.
Windows. I don't use it, and I have no clue how to debug installations on it. I haven't touched a dos command line for well over a decade. I suppose I could press a poor beleaguered university IT person into service, but that seems a little much.
Here are some alternatives:
Google Collaboratory
Pros: no setup, installation, or anything nasty like that.
Cons: Google. Who even knows if this will exist in a year given how often they kill projects. Essentially no tutorial material or UI afffordances, kind of ugly, have to figure out how to get code and data in and out. Requires constant internet access.
PythonAnywhere
Pros: established company with a good reputation, not going anywhere, beginner-friendly.
Cons: have to pay for anything that supporters Jupyter notebooks.
Roll a class server
Pros: complete control, no installation issues whatsoever, like PythonAnywhere but without charging the students.
Cons: way too much work.
The Harvard CS50 approach: create a class virtual machine
Pros: complete control over installation, can just distribute consistent version of every library up front.
Cons: a lot of work, students still might have installation problems with whatever virtualization option is chosen (vagrant? docker?), some students might not have machines with enough horsepower for virtual machines on top of everything else.
Just give up and use R
Pros: Better control over package version stuff, generally fewer glitches I think. Generally easier ecosystem for the kind of basic stats stuff students will be learning. RStudio is probably nicer than Jupyter Notebook on the whole.
Cons: R. I want students to see a general purpose, mainstream, programming language and get some idea of the things they can do with code as well as the stuff they can do with data. R is not even remotely suitable for general-purpose programming, as a language it has way too many warts. (1-based indexing? the evil that is stringsAsFactors? Ugh. Also, the library ecosystem issue.
What are your thoughts, world?