Notes from Bordeaux workshop

[jorisvandenbossche]

Notes from Bordeaux workshop:

• pandas intro notebooks can use a re-work:

  • data structures -> check the pandas-tutorial ones and compare (first dataframe, then series?)
  • data structures -> already do exercise with titanic? (loading the data, seeing the first 5 rows, plotting one column?)
  • basic operations -> countries["capital"].apply(lambda x: len(x)) is bad example, as doing apply(len) is the same
  • basic operations -> exercises with titanic ?
  • basic operations -> leave out alignment example? It only adds cognitive load, and does not occur in the examples (that's maybe for in the "advanced indexing" one?)

• pandas indexing / selecting data:

  • use titanic for exercises (instead of countries) ?
  • split in two parts? basic (selecting colum(s) + filtering rows) and advanced (actual non-default index, loc/iloc, assignment, index/multi-index)
    check: do we use loc/iloc in some of the case studies? (I assume we might use it for a combined boolean mask + column selection)

• new notebook: working with missing data ?

• reshaping: too much ? (only unstack and not stack?)

• bike_count case study:

  • don't use dayfirst as the solution, only show the comparison
  • more hints (eg "use value_counts", since they haven't seen that function yet)
    maybe work with hidden hint html?
  • drop_duplicates -> does not take into account the index!

• matplotlib visualization notebook:

  • ax -> ax1
  • set_axis_bgcolor no longer exists (g.ax_joint.set_axis_bgcolor('0.1') in one of the seaborn examples)

• general: more hints on what to use

• provide good cheatsheet

python intro: too little too fast if you don't know python yet, boring / too slow if you already know Python