The Fisher information is derived from the second moment of the log likelihood equation. I am unable to follow the step below (from wiki) moving from the Expectation of the second partial derivative to the (chain rule result?) difference of a second order partial derivative and the square of the first order partial derivative.

If someone were feeling "talkative" and would be willing to fill in the blanks there - it would be appreciated.


enter image description here


This is a chain rule problem.


d(log(x))/dx = 1/x  


 d(log(f(x))/dx = [d(f(x))/dx] / x

and: starting with the first order derivative:

enter image description here

Let us now take the second partial derivative by re-differentiating that first order one. This requires the chain rule:

d(f(a)f(b))/ dx  = d(f(a))/dx f(b) + f(a) d(f(b))/dx

Where we set

f(a) = 

enter image description here

f(b) = 1/ (f(X; theta)  = f(X;theta)**-1

The final result then follows.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.