Data Analytics and Visualization with R
- Individual repos called
dataessay_[GITHUB-LOGIN] with data and template document
- Repo should contain all material for replication, i.e. code and text
- Text should read like a paper, not a homework with chunks of code in-between
- Use chunk options to have a part with text only and code later
- RQ: To what extent is constituency work of legislative candidates strategic?
- Candidates’ motivations:
- Extrinsically motivated constituency work: Interact with constituents only since they could vote for me \(\Rightarrow\) responsive mostly to in-district emails
- Intrinsically motivated constituency work: Interact with constituents because this is an important part of my job \(\Rightarrow\) responsive also to out-of-district emails
- Legislative candidates are more likely to respond to emails from same district.
- Legislative candidates are more likely to respond to emails from same district, especially when candidates’ electoral safety is low.
- Data from field experiment from 2015 election of the Swiss legislature
- Random sample of 660 candidates across cantons and parties
- Researchers randomized two conditions in the email:
- in- vs. out-canton
- same vs. different party to candidate
Where To Start
- Get to know the dataset
- Start with a DAG (what is the main relationship? what are the confounders? what is randomly assigned, what is not? what is observed, what is not?)
- Choose your identification strategy
- Decide on model specification; have one main model (potentially testing all hypotheses), use it to generate meaningful quantities
- Present your results not only numerically but also graphically
- If necessary/possible, try different approaches for identification/measure for key variables/etc. as robustness checks
- No collaboration at this stage
- If you have questions/get errors and Google can’t handle them, ask me
- Deadline: June 13, 2023, 23:59
- Please upload PDF with signed statutory declaration on ILIAS
- All material (including the
qmd and rendered PDF) should be on GitHub by then, too. I should be able to render the same PDF with one click.
- Please proofread.