Read on Twitter
Who of the thousands of people who retweeted or liked that Bloom et al. attack thread has read their response?
It appears to show that neither censoring per se (but rather dropping half the obs) nor 1+patents materially affects results. https://nbloom.people.stanford.edu/sites/g/files/sbiybj4746/f/cm_response.pdf
It appears to show that neither censoring per se (but rather dropping half the obs) nor 1+patents materially affects results. https://nbloom.people.stanford.edu/sites/g/files/sbiybj4746/f/cm_response.pdf
It's easy to make sensationalist claims about a top paper to have been wrong, even when using doubtful techniques. ("It is usually possible to choose a subsample of the data to generate a negative result.")
Combine that with a conspiracy theory about top 5 journals, and you have a viral tweet.
It's harder to get people to look at the actual facts and form an own opinion.
It's harder to get people to look at the actual facts and form an own opinion.
Perhaps it's a good idea that the top journals don't generally publish bad replication/robustness studies, but only if there is an actual error.
Perhaps checking obvious robustness issues is what we have the review process for.
Perhaps checking obvious robustness issues is what we have the review process for.
