Thread by @SlickRockWeb

Last night we took a look at the Twitter dataset from suspended Kremlin troll accounts archived & posted by #NBC & look at the frequency of tweets discussing specific US states & display by swing states in 2016 election #Mueller Internet Research Agency nbcnews.com/tech/social-me…

As you can see there was a pretty clear grouping of purple states. This was an initial quick and dirty analysis and it doesn't take into account that bigger states would of course likely have a higher # of random tweets about them.

So in this next graph we normalized the # of tweets mentioning each state to the states electoral number thus normalizing the tweet frequency to the weight or importance of the state in the 2016 elections. Again a good grouping of purple states. #mueller Internet Research Agency

So normalizing for electoral importance, New Hampshire got twice as much attention as Texas did by the Russian twitter bots and Iowa got almost 3 times as much attention. I'm sure this had nothing to do with US politics (snark added).

This analysis can be further refined to look at a subset of the sample during the more appropriate last 3 months of the election & initial indications are its even stronger indictment that these Russian Twitter bots main goal was meddling in 2016 Presidential election

So I finally had a chance to read the Mueller indictment and I see now why my data is so relevant to today's events. On page 13 .."they learned they should focus on purple states like Virginia, Colorado and Florida". #mueller

For those that want to replicate this data. We used following regex 2 eliminate twitter handles that might include a state name: ( |"|')statename
We used the electoral numbers from the 2016 Presidential election. For simplicity we also did not include every state in the graphs.

Just noticed we made a mistake on the second graph normalized to electoral weight. Pennsylvania should be blue and not purple. At the time it wasn’t considered a swing state.