Final Peer Review #14
Description
This project aimed to predict the outcome of swing states in the 2020 presidential election. The group focused on 12 swing states, which were chosen based off 2016 election information. The data set includes many features, such as education, racial composition, economic features, etc. on a county level basis.
I really enjoyed reading this project. It was interesting to see how the group approached the problem. All the graphs and visuals were laid out well; these helped the reader better understand the data at hand and some of the transformation made, especially in terms of using the log county level GDP. I also thought that the group did a good job explaining their approach and how they handled missing data and how they split the 2020 data into a large and small set, as well as how they decided to normalize features or drop some that they felt would not add value. The actual explanations of the results were helpful, and I think these explanations helped facilitate the discussion about WMD. Additionally, the exploration of the mail in ballots was particularly compelling. Especially during this election year, I think mail in ballots drew a lot more attention. It was very interesting to see some of the visuals the group provided, especially in terms of exploring if Democrats were more likely to vote by mail.
One aspect that I think could have made the actual results more readable would be creating a table of the models and errors, rather than writing out sentences for all of these. I also think that it would have been interesting if the group explained the Bayesian Ridge Regression used to impute missing values; the group explained it briefly, but it would have been interesting to perhaps include a little more information regarding it. Additionally, perhaps since it was difficult to predict the vote share, it might have been worth it to explore just predicting if a county votes Democrat or Republican; in this sense, the problem would have become a classification problem, and it could have been interesting to compare their current model to the results of the classification model.
Overall, this project was very interesting! Especially after listening to our post-election lecture, I think that there is a lot of uncertainty and unmeasurable factors to consider for elections. However, the group did a good job trying to work with the available data.