I am self-studying the basics of logistic regression. I came across this sentence:

In logistic regression expected and observed information matrixes are equal

I am aware that the information matrix is part of the MLE, namely the second derivative of the log likelihood function.

I came across that statement when I was trying to understand the difference between Newton-Raphson algorithm and Fisher scoring algorithm as optimization algorithms. In Weibull distribution the optimization algorithms are necessary to find the maximum. This leads to my another question:

Why MLE for Weibull distribution cannot be determined analytically?

  • $\begingroup$ Can you please provide the source of that statement? (I can "almost readily" accept that they are assumed to be asymptotically equivalent but I am not certain about "equal") +1 Nice question. $\endgroup$
    – usεr11852
    Commented Aug 26, 2018 at 13:35
  • 2
    $\begingroup$ These are really two different questions, and you should split them and post them separately! $\endgroup$
    – jbowman
    Commented Aug 26, 2018 at 17:09
  • $\begingroup$ @usεr11852 i got it from my professors script. It is within the context of MLE in Beneralized linear models, where asymptotic normality applies. $\endgroup$
    – user1607
    Commented Aug 26, 2018 at 18:41
  • $\begingroup$ @jbowman i though there is a link, fixing it now then. $\endgroup$
    – user1607
    Commented Aug 26, 2018 at 18:42