Here is a sampling of my data projects, some completed through work and others as side experiments. Repos are either on my personal page or the Pew Research Center Github. Any side projects are my own and do not reflect research interests of the Pew Research Center.
- We manually collected the Twitter account handles of European political parties, then used the Twitter API to obtain account information for each.
- Then, we combined social media metrics with survey data related to party ideology. Using a random effects model, we predict whether more ideologically “extreme” parties have more Twitter followers.
- I presented this data at the 2019 AAPOR conference. It’s now published on Pew Research Center’s Medium site.
- After conducting 26 focus groups in the U.S. and UK, we wondered if we could use quantitative methods to more effectively analyze all of the transcripts.
- I used the
tidytext
package for basic text analysis, the stm
package to run structural topic models that compared topic prevalence by country and the tidymodels
and recipes
packages for a text-based classification model for topic prediction.
- Pew Research Center’s methods team released a package for survey analysis,
pewmethods
- In this post, I demonstrate the package’s capabilities for analyzing multi-country, weighted survey data with complex designs
- Additionally, I show two dataxiz examples with survey data using
ggplot
and rworldmap