Meta-trend analysis


  1. So far we have collected data for the past three weeks -- '200329', '200405', '200412' (for example, 200329 means we queried URLs collected from 03/22/20 to 03/29/20).
  2. The end query date is April 14, 2020 for all three datasets, because there werer some issues with URL formatting and I reran all analysis all that day. As a result, posts from 200329 dataset will last longer than 200405, which will last longer than 200412. This is an artifact of our data collection process.
  3. For each statistics, we look at how it changes as a function of time

Discoverability analysis (what percentage of URLs are searchable on each platform)

  • Facebook and Twitter consistently have high discoverability

User volume analysis

  • Among all sources, mainstream liberal is posted by the most number of users
  • If we look at fake sources, more users tweet conservative fake than liberal fake
  • Few verified users tweet fake source

Longivity analysis

  • longivity = number of hours between first post and last post
  • longivity has high variance, but overall (both in mean and median), conservative fake posts spread longer than liberal fake (this is true on all three weeks, and on both platforms)