Peer Review #4

[YawennnnnnTan]

Summary: Mingrui Li's paper "Predicting the 2024 U.S. Presidential Election" uses a poll-of-polls approach to estimate that Donald Trump will receive 49.7% of the popular vote. A linear regression model with three predictors—poll reliability, sample size, and pollscore—is used. The paper notes the model's simplicity but highlights limitations like the lack of state-level polling and potential biases. Improvements such as adding electoral college projections and addressing polling biases are suggested.

Strong positive points:

1. The poll-of-polls method is easy to follow and understand.
2. The paper discusses limitations, which shows an objective and self-critical perspective.
3. LLM usage is clear

Critical improvements needed:

1. Commit messages not informative
2. Scripts lack comments for the codes to explain the code
3. Analysis data is not in parquet format
4. Subtitle does not contain main finding, such as Donald Trump will receive 49.7% of the popular vote based on linear model with poll reliability, sample size, and pollscore variables
5. Lack of abstract (4 sentences) and repo link in the paper
6. Introduction section does not include enough background information. And cross reference is not complete when introduce the content of each section in introduction.
7. Lack of Data section which contains description/table/graph of raw data and analysis data. And lack of reason that when we choose poll reliability, sample size, and pollscore as variables in analysis data.
8. Model section does not include discussion underlying assumptions and model validation and checking procedures(such as plot of residuals vs. fitted value). And output about summary of model should not be directly showed in the paper, maybe form it as a table.
9. Lack of result section to show the model that you find and explain the model.
10. Limitation section maybe need the support of literature about how state-by-state polling and poll-bias include the result, to make paper more reliable.
11. Format of Appendix is wrong (please check the starter_folder) and Reference section is not in an independent page.

Suggestions for improvement:

1. Change commit messages to make them more informative, add comment to the scripts to explain the code and change analysis data format to parquet format.
2. Include the main finding in the subtitle and make the title more specific, add a repo link at the beginning of paper and add abstract (4 sentences).
3. Complete cross reference, especially in introduction section, and add more background information in introduction(3-4 paragraphs), maybe find some news or article as reference to make your paper more reliable.
4. Add description of raw data and analysis data, maybe add a brief table of dataset or graphs related to the dataset to provide your understanding of data. And explain why you choose poll reliability, sample size, and pollscore as variables in analysis data.
5. Add discussion of underlying assumptions and model validation and checking procedures, maybe in Model section and add result section to show your model and explain the model.
6. Change the format of Appendix be correct and start a new page for reference

Evaluation:

• R is appropriately cited: 0/1 use Python, but no python citation
• Data are appropriately cited: 1/1 cite the data properly.
• Class paper: 1/1 does not look like a class project.
• LLM usage is documented: 1/1 clear state in README and usage.txt.
• Title: 1/2 subtitle does not include main finding
• Author, date, and repo: 1/2 no repo link
• Abstract: 0/4 no abstract
• Introduction: 1/4 too simple and vague, should include more background information
• Estimand: 1/1 percentage vote is clearly estimated.
• Data: 1/10 simply introduction 3 variables but no data section and no graph/table of data and no reason of choosing variables
• Measurement: 4/4 description is clear
• Model: 2/10 no discussion of underlying assumptions and model validation and checking procedures
• Results: 0/10 no results section
• Discussion: 6/10 could be more deep and could use literature to support when discussing the limitation
• Prose: 4/6 good writing skills but paper is not complete.
• Cross-references: 0/1 incomplete cross reference, especially in introduction section
• Captions: 0/2 no table/graph lead to no captions
• Graphs/tables/etc: 0/4 no table/graph in model section, output is directly showed out, instead of forming as a table
• Idealized methodology: 10/10 The methodology is comprehensive.
• Idealized survey: 4/4 Construction and discussion about the survey is completed.
• Pollster methodology overview and evaluation: 10/10 really clear and in-depth and discuss strengths and weaknesses
• Referencing: 3/4 should start at a new page
• Commits: 1/2 should be more informative
• Sketches: 0/2 no sketches
• Simulation: 0/4 cannot find in scripts folder
• Tests – simulation: 0/4 cannot find in scripts folder
• Tests – actual: 0/4 cannot find in scripts folder
• Parquet: 0/1 wrong data format
• Reproducible workflow: 0/4 not reproducible, because path of getting data is from the local
• Miscellaneous: 0/3

Estimated overall mark: 52/126

[RohanAlexander]

@YawennnnnnTan - this is a perfect review. Thank you. May I use it as an example please?

[YawennnnnnTan]

Thank you. Yes, you can certainly use it as an example.

获取Outlook for Androidhttps://aka.ms/AAb9ysg

---

From: Rohan Alexander @.> Sent: Wednesday, October 23, 2024 8:47:14 AM To: aabbmddcc/US_election_prediction @.> Cc: Yawen Tan @.>; Mention @.> Subject: Re: [aabbmddcc/US_election_prediction] Peer Review #4 (Issue #4)

你通常不会收到来自 @.*** 的电子邮件。了解为什么这一点很重要https://aka.ms/LearnAboutSenderIdentification

@YawennnnnnTanhttps://github.com/YawennnnnnTan - this is a perfect review. Thank you. May I use it as an example please?

― Reply to this email directly, view it on GitHubhttps://github.com/aabbmddcc/US_election_prediction/issues/4#issuecomment-2432042975, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BLICZC353RTBGPWVJAITE7LZ46LFFAVCNFSM6AAAAABQNY3DFWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIMZSGA2DEOJXGU. You are receiving this because you were mentioned.Message ID: @.***>