amarchenko26
commented
5 years ago (4b) - if the paper said sentiment mean is -1, 1 it's either a mistake, or something's wrong in the code. i'm inclined to believe it's a mistake.
there's a merge conflict in 2_city_specific_clean.do, so i'm going to close this pull request - @ThiagoSchmidtResende can you pull the latest, figure out the conflict on your end, & pull request again?
… sentiment_mean
(1) Edited 2_city_specific: (1a) Melody removed the N of Obs bug. It was the drop if sex_res > 2 bit (1b) Removed [probably inconsequential] bugs
(2) Created review_cleaner: It allows review.do to work correctly. Backstory: review.do was originally not working properly because it one of its final steps was linked to the 1_cleaning file, which when 'done' would delete all of the code performed in the review.do file.
(3) reviewers.do: (3a) removed excess code, placed it in reviewers_trash.do, (3b) Identified a potential issue with sentiment_sd, but it turns out we do not use need that variable anyway. (3c) Identified difference between joinby and merge: merge is just a bit more flexible and transparent in terms of what data was not merged.
(4) reviewers_sentiment.do: (4a) Cleaned to specifications of previous regression files (4b) Identified issue with main variable in use: sentiment_mean_stan. No solultion has yeat been found. NB the paper said sentiments were judged from a [-1,1] basis. I have to look at R code to understand why sentiment_mean is not bounded by these values. I would assume its been normalized at some point. However, Regression tables are not the same when using sentiment_mean.
(5) summary table 4 begun: I began writing the code for the table.
(6) documents and tables.tex now have table 11!!