Code and packages

Analyis of Web Browsing Data: Practical Code Guide: This accompanies the paper published in the Social Science Computer Review and guides researchers through four steps of using browsing data (pre-processing, filtering observations, classifying, modelling). It will be continuously updated. Code

WebtrackR package: As a collaborator on this package, I am working on functions for preprocessing and analysing web browsing data, such as extracting components of URLs, estimating the duration of page visits and classifying visits. Code

Parsing Factiva output: Factiva is a great resource for working with past news articles. Which does not mean it's practical to use. Exporting search results to some processable format like csv is tedious: Your options are a PDF (not structured well enough for online converters), an RTF (again no luck with online converters) or HTML. This R function does the job of parsing Factiva results into a data frame. Code

Python function for approximating following dates of Twitter followers: The Twitter API (now defunct!) does not provide information about when a Twitter user started following another. This Python script allows you to infer the earliest possible date a user can have started following another user, exploiting the fact that Twitter orders the list of followers according to date of following. Code

---

Interactive apps and visualizations

"Visualizing regression" (Shiny App): One-predictor regression may be intuitive enough, but what about a model with several predictors and interactions? This app offers an intuitive understanding of what "ordinary least squares" does, by showing how it fits a line or plane to the data points. App | Code

"The role of randomization" (Shiny App): Experiments are considered the "gold standard" of causal inference: Through randomization, treatment groups are in expectation equal on known as well unknown variables. The app illustrates this by allowing users to re-assign treatment groups and observe group differences on a number of variables such as age and gender. App | Code

---

Data sets

"A list of over 5000 US news domains and their social media accounts": A list of US news domains with their associated social media accounts and ideology scores from Robertson et al. (2018). Data set | Release

"A list of Polish news domains with ideology scores and social media accounts": A list of Polish news domains with their associated social media accounts and ideology scores, estimated with a variant of the mediascores model. Data set | Release