# Thread: A problem of maximum likelihood estimator

1. Hi, everyone. I have a problem when reading the Population Genetics. The problem is as below:
P and q are the possibility of two alleles. The possibility to get a sample that contains G individuals, with a AAs, b Aas, and c aas is
[G！/a!b!c! ](p2) a (2pq) b (q2) c = [G! 2 b /a!b!c!] p 2a+b q b+2c
to estimate the parameters of the sample, we have the log likelihood function:
L = (2a + b) ln p + (b + 2c) ln q + constant
dL/dp = (2a + b) / p  (b+2c) / q = 0 ;
∴ p = (2a + b) / 2G
1/ V (p) = -d2 L / dp2 = (2a + b) / p2 + (b + 2c) / q2 = 2G/ pq. (1)

What confused me is that why the second derivative of the log likelihood function equals to the reciprocal of the variance of p?  2.

3. I'm not very good at this sort of stuff, but i can give you some terms that an internet search might help elaborate on.

the equality only occurs when the estimator (maximum likelihood estimator in this case) is "efficient" (term 1), for cases where the estimator is not efficient refer to the "Cramer-Rao Inequality" (term 2). Both of these cases assume that there is no bias in the estimation of the mle.

I'll dig out my notes from last semester to be more specific if this hasn't helped or if no one else can explain.  Bookmarks
 Posting Permissions
 You may not post new threads You may not post replies You may not post attachments You may not edit your posts   BB code is On Smilies are On [IMG] code is On [VIDEO] code is On HTML code is Off Trackbacks are Off Pingbacks are Off Refbacks are On Terms of Use Agreement