A thread for ASGN 2 clarifications and questions

[donbowen]

The assignment is live. Start it by clicking this link.

Good luck @LeDataSciFi/classmates-2024 !

1. Reply below with any questions you have!
2. The due date is the one on the schedule.
3. This is an individual assignment and any collaboration is forbidden. One of the TAs key jobs is to check for plagiarism.
4. Lehigh’s academic integrity standards remain the same whether you are learning online or in a face-to-face class. Any suspected violations of the code of conduct will be reported and handled through the University Student Conduct System.

Some clarifications on A2:

• The description of the dataset and variables is here.
• import statements: see comment below from the TA
• Q1: Use the natural log (this is what numpy does by default)
• Q1: You only need to report descriptive stats for the variables you created. You can do so in separate commands/outputs for each variable but I think it's easier to do all at once, and definitely easier to read that way.
• Q2:
  • Do EDA on the whole dataset. You can use eda.py to help there, and if you use it, do not put it in the gitignore. It should be synced!
  • See new comment below from me
• Q3: see comment below from the TA
• Part 4a: see comment below from the TA
• Q4: I ask about "the number of observations" and "the number of unique firms". Shouldn't these numbers be the same? No - those numbers can be the same, but they don’t need to be. Depends on the data set
• In part 4, I mention that I get N firms and had X observations with stock returns. That is meant as a checkpoint for you to match. If you get the same number of obs when you output the descriptive stats on the return variable, you're good.
• The instructions for using insufficient_but_starting_eda are here.
• Q5: It's asking for which is the highest average capx ratio. So take the average capx ratio for all the firms of each type, and output that. Which is highest?
• Q10 is a grpah with 2 lines, and Q12 is a graph with 4 lines

[BrooksWalsh]

Some Potential Clarifications:

1. Before Q1, in the import statements:

• from eda import insufficient_but_starting_eda
• You can delete the from eda import insufficient_but_starting_eda (or comment it out)
• If you want to use it, the community codebook has instructions about how to copy the eda function into the assignment. Don't gitignore it.

2. In instructions for Q3:

• 'Hint: firms_df[['leverage','high_leverage']].count() should produce the same number of non-missing values for each variable.'
• This should say "subsample[['leverage','high_leverage']].count() should produce the same number of non-missing values for each variable.
• Why? This hint is only true after filtering down to only gsector 20 & 45.
• Why are we clarifying this?
  • In the overall firms_df, there are 7 firms that have a defined 'leverage' value but DO NOT have a defined 'ind_med_lev'
  • Thus, if this hint is checked before filtering down, the count() comparison will be off by 7
  • These 7 firms have NaN as gsector, so it doesn't matter after filtering
  • In the overall firms_df, there are some observations with missing leverage but non missing ind_med_lev.
  • Weird! But common! Missing values are a fact of life with real data. We will learn more about them as we proceed.

3. Part 4A hint:

• The hint says 445 firms should download successfully. The true number should be 345.

Smaller, wording-based clarifications:

• Part 4b: Q9 says 'repeat for all portfolios'.
  • There are only 2 portfolios in the toy dataset. This wording is to prompt you to think about how you could do this for many portfolios at once if there were more than 2.
• Part 4b: Q10 says 'plot the 4 portfolios'.
  • Q10 is meant to refer to the toy dataset, which has 2 portfolios. You will be using 4 portfolios for Q11 and Q12.
• Part 4b: Q12 says 'adapt the code for Q12'
  • This should say 'adapt the code for Q10'.

[donbowen]

Assignment 2 due date updated to 8pm on Sunday

The schedule said 9am, I've adjusted it to 8pm.

@LeDataSciFi/classmates-2024

[elvinlee12]

@donbowen Can I copy code from the textbook? and should I cite it if I do?

[donbowen]

Yes you can and no you don’t need to cite

[patwilliams04]

Just a few questions regarding Q10-12. In Q10, did you mean to say “Plot the cumulative returns to the 4 *firms*”, as there are only 2 portfolios? And for Q11/12, what portfolios are you referring to? Should we assume you mean that we are to create portfolios based on the industry sector and leverage? Thanks for the help. Best,Pat

[BrooksWalsh]

Q10 is referring to the toy dataset, so for this question there should only be 2 portfolios. Q11 and Q12 are based on the dataset collected using yfinance in part 4a, and yes you should create the portfolios by grouping by sector and leverage, a total of 4 portfolios.

[elvinlee12]

@donbowen do we have to do q11 and q12? They aren't in the rubric.

[donbowen]

Yeah, it will be graded. I didn't update the rubric to reflect the change.

[donbowen]

Updated hint; please rerun your code today; remember to submit!

If you have already completed the assignment, please rerun your code so that you get updated answers, save, and push again.

This will help grading. (See below for why I'm asking you to rerun.)

Submitting your assignments

15 students have yet to push any work. Please do so! Remember, whatever is visible at github.com/ledatascifi/asgn-01-
is what you are graded on.

Updated hint

The assignment says this:

• HINT: Some stocks wouldn't download because of a problem with the ticker (usually they are delisted now). Ignore these and keep going. I was able to download prices for 345 firms and had 20,636 observations with a return value.

Those numbers were true last Sunday, but no longer. I just reran my code and it appears that 1 of the tickers no longer downloads from yahoo. Now I have 344 firms with return data, and 20,576 observations for the return variable.

@LeDataSciFi/classmates-2024

[hanadinh]

When I calculate daily returns and cumulative daily returns, I got numbers smaller than -1. However, when investing, you cannot lose more than 100% of the original amount. Should I just cap everything at -1 or ignore it?

[donbowen]

you’re right that this indicates an error in your code. Capping losses at -100% won’t fix it. In the toy example, all firms and portfolios add money. In the real data, the portfolios lose money but nothing near all.

Without seeing your code it’s hard to guess, but perhaps look at how your return variable is scaled. -10% should be -.1, not -10. (This happens sometimes when students copy too much of the other code.)

[donbowen]

Assignment 2 is in the books!

Hi all @LeDataSciFi/classmates-2024

We will talk about assignment 2 this week, but I want to post something right now.

Assignment 2 is usually considered the hardest assignment of this class. The right-sized goal isn't getting all questions correct, but trying hard, learning some ways to interact with pandas dataframes to do EDA steps, and getting used to finding resources on the website to figure out a step or two on the assignment.

I want to congratulate all who worked hard on this assignment (regardless of how many questions you got right). That level of effort is laudable and praiseworthy and nearly always ends in amazing learning outcomes. (@BrooksWalsh will attest to this.)

Looking forward:

• The next two assignments are generally regarded as a large step back in effort. We will continue learning new things, but at this point, we've gotten our feet wet on using github + python + pandas. That's a lot, and we jumped into the deep end, but the learning curve will soften from here.
• There are 9 assignments in the class, so the grade on each is relatively low weight, meaning there is plenty of time to make it up.
• If you're struggling or not: come to my office hours or send me an email to schedule a talk! I want to help everyone learn the material; I've had very little office hours traffic or emails. I can only help those that reach out.