4 ms·
I disagree. You could write personalized letters to each of thousands of professors, as long as each contained a slot for the student's name. If you then random
by emiliobumachar 13y ago
I disagree. You could write personalized letters to each of thousands of professors, as long as each contained a slot for the student's name. If you then randomly assign the different student names to the letters, you still have a perfectly valid controlled experiment. No need to compare two outreaches to each other.
Being said that, the study as it is is already very solid proof of discrimination. So what that the email looks spammy? If it looks just as spammy from each and every group, and not every group is equally rejected, that's textbook discrimination.
- thaumasiotes 13y agoThose personalized letters won't be comparable unless they are so generic that they're still spam. After randomizing which name goes to which letter, you're stuck with an infinite number of dimensions of variation, where one dimension is the ethnicity of the name signed at the bottom of the letter, and the other infinity represent the "quality" of the personalized letters. You won't be able to conclude anything about the influence of the name.
- emiliobumachar 13y agoI still disagree. If there are only two letters, than, sure, the relative quality of them will dwarf the influence of the name. But if there are thousands of letters, assigned randomly, the Law of Large Numbers does its magic. No name gets assigned significantly lower or significantly higher quality letters than the average. (something like that could happen, but the odds are small enough to be ignored). http://en.wikipedia.org/wiki/Law_of_large_numbers http://en.wikipedia.org/wiki/Law_of_large_numbers