The estimator (for the mean) of x-bar + 1/n is consistent, but biased. As n goes to infinity, 1/n vanishes, and x-bar converges to true mean. In finite samples, the 1/n term is the bias. So consistency does not necessarily imply unbiasedness.
At 1:19:11 you say that unbiasedness implies consistency, but you are wrong here, I'm afraid. All combinations of bias and consistency co-exist in fact.
Bias and consistency do co-exist, he clarified that point. But a universal affirmation doesn't convert universally. In other words, that does not mean that all consistency co-exists with biased-ness. In the second place, a truly unbiased estimator does imply consistency because at any point in time it will be unbiased -- which is the definition of consistency (i.e. robust against temporal intervals)
I thought of a heteroskedasticity case where larger x's produce lower error variance. Not an economic application, though. Take Xi = body weight of subject i, Yi = how high subject i can jump.
I have a question. In the lectures from the econometrics courses I have in college, we always see T tests with sqrt(Var(B|X)) why do we see it here always with sqrt(Var(B)) instead? being B the Beta. What is the difference? anyone? thank you
That book is garbage, I'm watching these videos exactly because the video is very unclear on key concepts and formulas and instead spends chapters on useless examples