It seems that Trump's tweeting was positively correlated with metaphorical violence usage in the media. I have found instances of the media reporting on Trump's tweeting, describing it in terms of metaphorical violence. For example, Trump urged readers to find Alicia Machado's sex tape and released a video that accused Bill Clinton of "assaulting women" and said Hillary Clinton tried to undermine the women's claims of assault.
So, it makes sense to try to quantify this relationship. Trying to quantitatively establish causation in this case may be beyond the scope of this paper. Establishing correlation is straight-forward.
In [1]:
import os
import warnings
warnings.simplefilter('ignore')
os.environ['CONFIG_FILE'] = 'conf/default.cfg'
from correlate_trump_tweets import correlate
In [2]:
ts_data = correlate()
# Uncomment below to persist data in a series of CSV files.
# ts_data = correlate(save_dir='data/metvi-tweet-timeseries')
In [4]:
ts_data['trump'].resample('1w').mean().plot(label='Trump tweets')
ts_data['clinton'].resample('1w').mean().plot(label='Clinton tweets')
plt.ylabel('1-week Rolling Average')
plt.xlabel('Date')
plt.legend()
Out[4]: