Tweets and Storms: Replicating Holler (2021) Study on Hurricane Dorian

In order to replicate the Holler (2021) Hurricane Dorian study with data from Hurricane Ida, I focused on adjusting the computational environment so that all packages were usable with the code from Holler (2021). Another goal of this replication was to adjust the normalization process to use Census data rather than X/Twitter data. This is due to increased boundaries to data from the current X/Twitter API. Additionally, there were a few unplanned deviations such as small visualization adjustments and data transformation. Lastly, I added a discussion and conclusion section after the analysis to help explain the context and implementation of my modifications.

Throughout this replication process, I have learned a lot while also reinforcing key concepts from this course (Open GIScience). Firstly, the complexity of maintaining computational environments is important to acknowledge going into any study. Specifically, the ‘rtweet’ package had changed since 2021, so Elise and I had to install an older version. Additionally, I also learned a lot about curating a research compendium. By writing out a fuller preanalysis, metadata index, and more, I felt more comfortable with the components of the study; I am also proud that my work is now more transparent.

I think this project showcases my growth and understanding of open GIScience. This replication is also an interesting case study for social media analysis during a turbulent time for the relationship between social media companies and researchers.

Check out the replicated study here

Check out the repository for the replication study using Hurricane Ida data here

Also, I would like to shout out fellow student Elise Chan for the help and collaboration while working on this project!

References

Check out the original repository for the Holler (2021) Hurricane Dorian study here