kbuzard
commented
3 months ago @ecn310/development Here's my feedback on the .md file that @spsotelo posted here early this morning.
Let me start by saying: this represents a ton of progress! Well done!
I'm going to divide my comments into those on process and product.
Process
- You need to translate this into Overleaf for the next time I look at it (see instructions for creating the overleaf document on Slide 10 on November 7). This makes it easier to give feedback on details, and is how the final report will be submitted, so it makes sense to iron out issues there early on instead of having to translate it later and having to deal with all the LaTeX issues then.
- This means that I'm only give broad feedback at this point; I will give you more detailed feedback when you've addressed this post and moved it over into Overleaf.
- I need to see your code that produces these outputs. It should be well organized and uploaded to this repo.
- Your code that produces the outputs should save the outputs into your Variable Graphs folder on Github.
- Then, when you sync to Overleaf, you'll have easy access to the figures you need to insert.
- I don't know where the regression results came from, as we haven't learned to use regression in class. You should stick to statistics we learned about, or talk with me if you want to use something more sophisticated.
- There should be something on your daily log about doing the analysis work and where to find it. This is separate from writing up the results, which you reported on well.
- If this came from ChatGPT or similar, please carefully read the item 3 in the list at the end of the syllabus and adjust accordingly.
- "To resolve this, I plan to use Photoshop to manually label the axes later this morning."--don't bother with this. I'm putting together the reading for next Tuesday, and it will be about how to do these kinds of things in Stata. We should definitely do it in Stata, because doing it in Photoshop would not be reproducible. I'll include something about adjusting the axes on the graphs.
Content:
- Your report indicates that you're working with aggregated data. I'm assuming is means that it's taking averages across countries for each year (you say "which may mask regional variations within countries", but I don't think you have sub-national data so I think you might have meant "which may mask important variation between countries" or something similar).
- This aggregate data approach can be useful for learning about your data, and may be worthwhile to present a little of this to give readers a broad overview in the data section.
- MOST IMPORTANT: Where is the data section? It provides important context for the results section, so it's hard for me to give detailed feedback without it.
- Your analysis should mostly be at the country-year level since averages can hide a LOT of what's going on. I think this is what you're acknowledging in the limitations section.
- This may help to explain another question I have: why does the graph for manufacturing employment in developing countries bounce around so much? Is it because some countries only have data for some years?
- This aggregate data approach can be useful for learning about your data, and may be worthwhile to present a little of this to give readers a broad overview in the data section.
- You present a t-test for the GDP being different between developing and developed countries. But this is true by construction/definition, so this is not useful to report. Better would be to discuss
- You say that the graphs exclude 2020; I think it would be better to just stop your analysis at 2019 to avoid issues with the pandemic.
- You should present p-values with the correlation coefficients (use pwcorr with the "sig" option like I showed you in class).
spsotelo
I've created this issue so there's a place to keep my feedback on your write up about your analysis (the results section of the report).