Hi, your videos are always the best. I was wondering: to get the upper bound and the lower bound is always +/- 1.96? Because you mention that that would be our critical value, but I cannot see that result in the regression run. Thank you very much!
I saw that dewey regression in a paper that I am trying to recreate and I have to say that it seems like bullcrap. I tried it on my data and nothing much changes, just a bit on the standard error on a couple of my variables but the estat bgodfrey test (or the dwatson) does not change for me. I prefer the other options.
I'll assume that you are referring to the Newel-West robust standard errors here. The idea of this adjustment is to calculate the variance and standard errors with a formula that does not require zero autocorrelation, while leaving the actual regression equation and coefficients unchanged. Therefore, the error is still autocorrelated (thus no change in your dw or bg stats), but the results are valid.
Thank you!! this is really helpful to my research.
great video, thank you!
Hi! Thanks for your video, it is really helpful. Could you do one where you use rolling regression with panel data? That would be wonderful!
Hi, your videos are always the best.
I was wondering: to get the upper bound and the lower bound is always +/- 1.96? Because you mention that that would be our critical value, but I cannot see that result in the regression run.
Thank you very much!
the critical value 1.96 correspond with 95% confidence bands
I'm very interesting in your video. Very helpful! Thanks a lot!
Can you show please out-of-sample data recursive regression and R-square value?
I saw that dewey regression in a paper that I am trying to recreate and I have to say that it seems like bullcrap. I tried it on my data and nothing much changes, just a bit on the standard error on a couple of my variables but the estat bgodfrey test (or the dwatson) does not change for me. I prefer the other options.
I'll assume that you are referring to the Newel-West robust standard errors here. The idea of this adjustment is to calculate the variance and standard errors with a formula that does not require zero autocorrelation, while leaving the actual regression equation and coefficients unchanged. Therefore, the error is still autocorrelated (thus no change in your dw or bg stats), but the results are valid.
thank you for this useful video. i have some questions, if you agree to discuss them
First on this!