Proposal feedback

[donbowen]

I think you have a nicely formatted proposal and it's a very financey question.

Your discussion of your testing metholodology and the data is a little off. Here are some things you should update:

• Your target variable isn't the sentiment score. It's the market and sector ETF returns. (So, you actually have a bunch of target variables.)
• Don't use returns just one week before and after

By making some small changes to your dataset, we can do your regression tests AND some nice graphs:

• Your dataset is a "tall" asset-date dataset. Like our beginning of semester stock return datasets. You'll add variables for:
  • Which is the nearest fed announcement (maybe some kind of "event id" variable)
  • excess return (ret minus market ret)
  • cumulative excess return for that asset during that event's time window (from t-10 up to t+10)
  • sentiment of the corresponding fed announcement
  • some variable(s) that cuts the sentiment variable into smaller bins (e.g. High/Low, High/Middle/Low)

I'll add a picture below to elaborate.

@SikaiWang224 @stevenwang1213 @lxl0816

[donbowen]

Below is a suggested plot for one industry. You could repeat this for each of your 10 industries.

One additional plot could have the cumulative excess returns for all 10 industries for all fed announcement. So 10 lines. That’s a lot of lines but in assignment 3 or 4 we made a nice version of a graph to display that info nicely.

One other plot could, for each industry, get the difference between the hi sentiment Cumret and the low sentiment Xumtet and then plot all 10 of those lines on one plot. That’s a lot of lines but in assignment 3 or 4 we made a nice version of a graph to display that info nicely.

The attached pic also has a mock up of the dataset and defines 2 variables. I provide hints for the plots and the regression.

[SikaiWang224]

Changes:

1. Change the target variables from sentiment score into market and EFT returns.
2. Add a piece of paper that includes a sample dataset includes the required variables like announcement ID and returns.
3. Add information that describe what is the structure of the team website.
4. Add the process of how to load our jupyter lab notebook.