I understand that the only way to ensure getting geocoded tweets using the twitteR package is by specifying a geocode. However, apparently, even if you specify a wide range (e.g., 3000 miles) priority is given to tweets near the center of the range. e.g.:
tweets<-searchTwitter('s', n=2000, geocode='37.000,-120.00,3000mi')
gives you mainly California and Arizona tweets. I don't see that feature/bug documented anywhere, but was a pretty robust finding when I tried lots of different spots (e.g., specifying the middle of the US gave mostly IL/AR tweets).
Is there any way around this to get mostly American geocoded tweets? I'm happy to do some filtering, but given the time to get tweets, I don't want to have to filter out 95% of the data.