kbjarkefur
commented
2 years ago This comment made me think we should have an advanced topic "Cleaning and analyzing tabular data". For that session you are expected to know how to load data in pandas, know what a series is, know how to do basic subsetting from the "foundations" part of the course, but then in that session dive into details.
weilu
My feedback for this session is still what I wrote in https://github.com/worldbank/dec-python-course/issues/25. That's why I did not add anything here during the session.
It really comes across how well you know this material @weilu , as you present it really well. But no matter how well anyone would present it, it is too much information for any human to take in. You need to already have a fair bit of experience in python to have a chance to follow along. The information was presented in the speed of a recap when the objective of the session is to be an introduction.
I know it is hard, especially when you have prepared really good material, but darlings need to be killed. One way to do so is to think about "do not give a man a fish, teach a man to fish". Listing this many things you can do with numpy and pandas becomes handing out a lot of fish. You did teach how to fish as well, but I think the participants was overwhelmed by all the fish to be able to be able to pick up on the lessons on how to fish.
To give a concrete example. I can never think of a situation where I needed to reshape a ndarray. I asked Luise and he said it sometimes is used for geodata. If we only can come up with needs for reshaping in special cases, then I think that should be taught in the advanced topics sessions where it applies and not in the "foundations" part of the course.
Suggesting slashing the numpy section into: 1. there is something called numpy and ndarray, 2. an ndarray is different from a list in this way. 3. Here are use cases for an ndarray. 4. in those use cases introduce max three things you can do to an ndarray.
That will leave much more time for pandas which is something much more general and should be given <80% of the time in a foundations course. Then, for example, have a full section of explaining boolean indexing where the data frame is small enough that you can show the boolean array next to the data frame so people understand what is going on. If people understand this well, then they have the foundations to be able to understand documentation, read solutions on stackoverflow etc.