![]() For 20 regressions, we expect to get one false positive: a result with \(p<0.05\). Here’s the basic idea: first, p-hack a significant result by running regressions with many different treatment variables, where the true treatment effects are all zero. In this post, I show that it can be easy to p-hack a robust result like this. The twist: I p-hacked this result, using data where the true effect of Because the coefficient on \(X\) is stable and significant across columns, we say that our result is robust. ![]() Economists want to show that our results are robust, like in Table 1 below: Column 1 contains the baseline model, with no covariates, and Column 2 controls for \(z\). ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |