Statistical modeling project supervised by Andrew Gelman, in partnership with YouGov.
We're analyzing a survey dataset of nearly 6000 respondents. In addition to standard political preference and demographics questions, people were asked to describe their friend groups in terms of party affiliation (Mostly Democrat, All Republican, etc) and wealth (All less wealthy, Mostly more wealthy, etc).
We plan to analyze this data using multilevel regression, to understand how these social variables relate to people's voting preferences. Additionally, we'll analyze how people's views on major political issues have changed throughout the past few months—and especially how those changes vary by social situation. For example, we might ask how Democrats with majority Republican friends living in swing counties have or haven't changed their views on Social Security.
See the /notebooks subdirectory for knitted notebooks with exploratory analyses of these friend_group_copartisanship and friend_group_class variables.