# exp R1 distribution of tweet types with matching domains ```r type n_tweets pct_tweets <char> <int> <num> 1: retweet 2428321 89.35 2: tweet 155343 5.72 3: quote 90078 3.31 # quote tweets 4: reply 37190 1.37 5: quote-reply 6853 0.25 ``` modeling/results: quote vs. retweet t-values (for winsorization = 0.95) - quote-tweet results are directionally same as retweets, but t-values are less strong (ie smaller effects) ![[compare-with-retweets_tvalues_conditionC.png]] cluster by user vs block - specs where user SE<=clustblock SE: 162/162 (varied winsorization, outcome [count/sum/frac], threshold) - clusterblock is more conservative (wider SEs)