Testing for endogenous instruments - test for overidentifying restriction
HTML-код
- Опубликовано: 11 сен 2024
- This video outlines how the test for endogenous instruments works in practice. Check out ben-lambert.co... for course materials, and information regarding updates on each of the courses. Quite excitingly (for me at least), I am about to publish a whole series of new videos on Bayesian statistics on youtube. See here for information: ben-lambert.co... Accompanying this series, there will be a book: www.amazon.co....
Hey Ben, I'm an economics student in college and your lectures on the various tests have been very helpful. Thanks!
Your explanation is so spot on.
For the test of over identification at stage one you stated that you estimate the equation by two stage least squares to get the estimated values of alpha and beta. Did you mean OLS? The part I am asking about starts at 3:48
+jods kin I don't think he meant OLS, you use two stage least squares to get rid of the endogeneity part in X. You do this by regressing X on z, to get values of alpha 0 alpha 1 z, this new value of X will now be estimated. So in order to keep notation clear, putting it back into the correct form, it is y = gamma 0 + gamma 1, because now your new error has values that werent previously there, so you have a new error too.
this is where the notes on board are confusing. The residuals are taken from the second-stage regression of the 2SLS.
@@zjli1234 Can you explain? I think 2SLS might be correct, though it is not the same as the usual one. It only is used because there are "2 stages", namely first OLS, then predicting the residuals with our IVs. Am I mistaken?
@@AloofMusician Are you sure?
This lecture is superb but would be grateful if labeled by lecture 1,2,3,.......... etc
see his playlists.
Thank you very much for sharing this!
I have a question: Why can we use this test for finding whether z is exogenous by regressing the estimated errors from the original regression on the variable z? I thought the estimated errors (or residuals) cannot be used to test the relationship between the actual errors and the variable z (exogeneity assumption). Would be glad about having an answer! Thanks!
he says "some part of the error". We cannot check if x is correlated with the error, but rather z and our estimated error, which in theory should be somewhat related to the true error. Not sure why though.
I love your videos, they are so helpful! However, this is the first video I am finding a little more difficult to work out. In our econometrics class, we say that if we have multiple instruments for one endogenous variable, the instruments will be considered relevant if the F-statistic from a joint test on the instruments is greater than 10. Is this true? You mentioned something about if the Rsquared is small, we will not reject the null... what do you mean by this?
The instruments will be considered JOINTLY relevant in the regression because the F-statistic is so ridiculously high (10) that you reject the null hypothesis, which says that they're not significant together (null hypothesis states that all their population coefficients are zero). Under the null hypothesis, the test statistic is NR^2 ~ chi-squared with some df. This means that for a high R^2, the test statistic will be higher and you will be more likely to be in the rejection region. This high R^2 also shows that a lot of the variance in the residuals can be explained by the instruments, and thus they are correlated (which is a no-no for instruments).
Ben, do you know if there is a statistical test for instrument strength and validity?
Thank you sir. It's super helpful!!😭
Thank you very much for the video. However, I have trouble understanding why this test requires two instruments. Any answers would be greatly appreciated! Thanks
Did you find the answer?
this is so helpful, thank you!!
I appreciate your lecture. It is very helpful. :)
Thank you Sir :)
3:58 you said estimate these 2SLS, did you instead mean via LS? Coz in the conclusion part you mentioned that 2 IVs were required otherwise epsilon hat would be baised
Do you know it now if it is LS or OLS?
@@lastua8562 www.econometrics-with-r.org/12-3-civ.html it is the error from 2sls.
i love you.
do you know how to do Hansen test in Stata? Thanks!
I know this method, but I don't think this method is good. If z1 and z2 are all bad IV, the residents of the IV regression will be not good alternative of the random iterms.