# Module 3 taught element feedback Delivered by James Robinson and Katriona Goldmann - We copied the delivery from previous year allowing students to comment on the 'figures gone wrong' section in a [hackMD](https://hackmd.io/PKxvJS4XS1O7tyLXOb0xIQ). This interactive element went down very well. Also allowed them to give examples of poor figures in the wild. - James and Kat delivered alternating sections with breaks in between. Overall stuck to timing quite well. - Walk-through example very useful to give students oppotunity for hands-on work. # Module 3 hands-on feedback ## Summary of task Participants are divided in groups of 4 to 5 participants and paired with a helper. They work in groups on one of the two suggested tasks (looking at relationships between variables or investigating missing data). Idea is that the participants decide what they think is interesting and go and do it. ## Comments - It would be nice if there are solutions - The session is meant to be free-roam, but it's probably _too_ free-roam in it's current state - Would be useful to have some guides for both the instructors and the participants benefit - Could argue that this could be solved if the instructors did go through the data and tried to do the task themselves, but it's hard to think on the spot about what are good avenues to go down if they didn't - Instructors probably don't have time to try the task themselves before the session anyway - Both instructors and helpers should be clearly told that delivering this session is quite a bit of work, they need to be already familiar with the dataset (which is overly and very unnecessarily complex), and they need to be pros at plotting with python so they can actually help during the sessions and not say something that would be wrong. I (Levan) definitely felt I was not that helpful since most of my plotting experience is in R. - These "solutions" don't need to be full blown notebooks that we share to the participants as "good data exploration", but even some example plots that we can suggest the participants to try make - Or some interesting questions (that are bit more specific) to get the participants to do - Could just make the task include more proposed questions - Participant also asked too for solutions (both for hands on session and earlier taught session) - The suggestion of working on missing data methods (imputation, etc. in module 2) is really unrealistic - The complexity of the dataset can be problematic: - Students spent long time familiarising selves with dataset and understanding what might be good to explore ergo. less time playing around with plotting and different visualisation techniques - Potentially provide a subset or simplified version of this data for the hands on session - Difficult for instructors who haven't been involved in earlier sessions to jump in midway without spending a long time preparing for session and going through the data - Might be resolved by providing solutions/ more structure for the session as mentioned above - Originally the plan was to try to tie together the hands-on sessions more closely so e.g. Module 3 would use slightly pre-processed data already prepared during Module 2. That would be a reasonable amount of work to do (and we ran out of time to do it that way when developing the course), but could help with this. - How to collaboratively work on a notebook? This was something discussed by instructors as multiple groups were unsure how to do this (as were we) - VS code? - Generally people worked by having one person working on notebook but sharing their screen - [name=Levan] Perhaps people do not need to work collaboratively on plotting and exploring the data? Is that a realistic scenario, where group of 4-5 people together have to plot and create a presentation? I think it would be best if everyone is allowed to do their own work, but have others in their breakout room as support or source for inspiration. - Windows support (not really a point specific to this session) - I feel we don't really support windows users very well. I (Ryan) had to help someone using Windows and it wasn't a hard problem but it wasn't very smooth - Not sure the solution to this though. This has been a problem in RSE course and other teaching stuff we've done in the past - Potentially better to move away from using poetry and instead use conda (students are more familiar with this set up and is easier for Windows) - I agree, and there's an old issue about this: https://github.com/alan-turing-institute/rds-course/issues/112 . It adds unnecessary complexity when we're not trying to teach about environments etc. in this course.