0

There is this rule of thumb that the first stage F-statistic should be >10 in instrumental variable analysis to rule out weak instruments.

Is this "rule" for one instrument only or also valid if I have many instruments?

spore234
  • 1,323
  • 1
  • 15
  • 31
  • What do you mean by "many"? Are you in a high-dimensional setting? – Andreas Dzemski Apr 09 '19 at 13:57
  • @AndreasDzemski as many as 10 or maybe 30 – spore234 Apr 09 '19 at 14:33
  • See this answer: https://stats.stackexchange.com/a/184803/241093 – AlexK Apr 14 '19 at 03:41
  • @AlexK so this rule of thumb is also valid for more than one instrument? The answer is only about one instrument. – spore234 Apr 15 '19 at 12:49
  • 1
    The "joint significance" in the first sentence of that answer relates to the general case of more than one instrument. F-test tests the joint significance of all your instruments, with the null hypothesis that all of them are uncorrelated with your endogenous explanatory variable. – AlexK Apr 15 '19 at 19:23
  • Is that what you meant http://www.mit.edu/~hull/CIV.pdf ? – Sergey Aug 13 '21 at 00:26

0 Answers0