key Jupyter notebook mods

[L4R5m]

Changes to key.ipynb:

• Add section headings and Table of Contents (visible if you have ToC2 nbextension installed and enabled)
• show how to set Pandas options
• refactor and bugfix data wrangling code
• update matplotlib API example to match R ggplot example

[L4R5m]

I read Sam's presentation and now see that we want to convert the dataframes to 'tidy long form', not just find the quickest / easiest way to calculate 'number of colors' & 'number of description words'. So let me have a look at doing that... your original solution may have been the right approach!

[seanhelvey]

I'm still happy to merge this if you'd like @L4R5m. I really like the comments, headings, and formatting. The big challenge is that Sam and I were trying to follow a very simple format of 1) read the data 2) wrangle it 3) visualize so that newcomers would be able to follow. I think we need to distill this so that it maps more closely to Sam's solution here https://github.com/samanthacsik/RLadiesSB-RvsPython/blob/main/KEY_Rcode.R. Can we get it down to three simple steps?

[L4R5m]

Sounds good, I updated the Notebook to remove one of the analysis methods so now it is just based on the 'tidy' dataframe method. Feel free to edit further if you want to simplify more.

[seanhelvey]

I don't think I can modify the pull request since you forked @L4R5m. I did my best to incorporate your code into key-2. @samanthacsik and @an-bui feel free to give feedback here too. Some notes:

• I've gone back and forth between using a lamba vs Tidy format one per row. Why don't we do colors Tidy style and then use a lambda on the image description so they can see each?
• We can make it very simple and clear for attendees to get into the notebook and write code by starting with the pandas steps 1-3, and then introduce Matplotlib and NumPy afterwards.
• Since we are going to be writing every line of code together for the first few steps at least, we can eliminate the formatting options by not using print statements, so output is nicely formatted by default.
• I am conflicted between filtering using df[df['hs_tf'] == 'Yes'] and df.loc[df.hs_tf == 'Yes'], but to me the first seems easier to understand, even though I know the docs say .loc is best.
• Maybe we can avoid dot notation and just use square brackets for consistency?
• Is the sort=False in the group_by for performance? I'm trying to eliminate anything we can avoid explaining.