Feedback on init proposal

[donbowen]

@Danielshin2002 @MariaMaragkelli @justinreed23 @reghanhesser

Really nicely written! Gave you a 100. I have given you a lot of offline feedback (the most of any group), so I'm going to keep this brief and fast:

• For the revision, focus on adding this to the outputs section: I'd like to see mockups of what your dashboard will look like and show. You can hand draw things and upload the a pic (ok) or use the drawing tool mentioned in the discussion board.
• Your data that you save should just be 1 CSV file with ETF, date (month), and return. No need for one csv per ETF, no need for price vars, etc. Not clear if you were going to save 1 or many. Students often save 1 per ticker, not sure why.
• return should use Adj. Close.
• Your fake dataframe is a little too simple :) (If you'll have multiple ETFs, your example df should have a few rows from each, which makes clear that the dataset contains multiple ETFs.)
• An issue you referred to: Your ETFs will not have a long enough data span to build out 40+ years for your simulation. There are a few ways to deal with this. The simplest is to sample from your existing dataset, which will preserve correlations between the assets:
  • Reformat your tall dataset ( ETF, date (month), and return) to a wide dataset (month, then columns for the return of each ETF). See the textbook for how, but something like wide = df.set_index(['date','ticker'])[['ret']].unstack() should do it.
  • Drop rows (months) where you don't have returns for all the ETFs you want. You need to have rows where all returns are available.
  • Randomly pull rows from this dataset until you have 1280 months of data. `wide_simulated_life = pd.sample(n=1280, random_state=42, replace=True)`. Replace=true allows you to sample the same month more than once, so that wide_simulated_life can be bigger than wide.
  • Save that.
  • You can wide_simulated_life into tall_simulated_life by unstacking. Refer to textbook.

[donbowen]

Whoa! This is dangerous given your ETF choices: "We will drop all rows that are missing at least one ETF"

Think about it - one of your ETFs started in 2022, another just in 2018. So really you'd be doing this just on 2022-2024 data. My suggestion: replace those ETFs with longer lived ETFs or delete them.

@justinreed23

[donbowen]

Your final dataframe won't have the real dates, it should have fictional dates, so I suggest it be month 0, 1, 2, .... 600.

Should the mock up have a savings rate? ($ per year or % of gross income?)

Should the output highlight also the line corresponding to the portfolio their risk aversion or whatever would suggest?

Nice mock up @Danielshin2002 @MariaMaragkelli @justinreed23 @reghanhesser