My Math Forum  

Go Back   My Math Forum > College Math Forum > Advanced Statistics

Advanced Statistics Advanced Probability and Statistics Math Forum

LinkBack Thread Tools Display Modes
May 10th, 2017, 11:19 AM   #1
Senior Member
Joined: Feb 2015
From: london

Posts: 121
Thanks: 0

Likelihood function

Group A - women with no children
Group B - women with children, where number of children is distributed with poisson with mean u

Probabilities of being in group A and B
P(A) = p
P(B) = 1-p

$\displaystyle \phi =(p,u)$
$\displaystyle L(\phi;y) = [p + (1-p)e^{-u}]^{y_0} [\frac{(1-p)e^{-u} u^1}{1!}]^{y_1} [\frac{(1-p)e^{-u} u^2}{2!}]^{y_2} ...$

where $\displaystyle y_k$ denotes the number of women with k children

Please can someone explain how this likelihood function is constructed. In the notes it says we can think of this problem as a mixture of a Bionomial and a poisson distribution.

The thing that is really confusing me, is why is each section to the power of $\displaystyle y_k$
calypso is offline  
May 11th, 2017, 12:37 PM   #2
Senior Member
romsek's Avatar
Joined: Sep 2015
From: CA

Posts: 1,199
Thanks: 611

I assume you got as far as

$p_k = P[\text{a woman has k children}] = \begin{cases}

p + e^{-u} &k=0 \\

(1-p)\dfrac{u^k e^{-u}}{k!} &0 < k


We sample a bunch of independently selected women and note the number of children they have.

The joint probability of the number of children of all these women is just the product of all their individual distributions above.

Now you coalesce these observations into a count of how many women have $k$ children.

To try and save me some typing imagine that we have a product of different $k$s

$p = k_1 k_0 k_1 k_2 k_3 k_0 k_1 k_4 k_3 \dots$

we can rewrite this product as

$p = (k_0 k_0)(k_1 k_1 k_1)(k_2)(k_3 k_3)(k_4) \dots$

$p = \prod \limits_{j=0}^\infty k_j^{\text{# of women w/k children}}= \prod \limits_{j=0}^\infty k_j^{y_k}$

Now just replace $k_j$ with $p_k$ above

$p = (p + e^{-u})^{y_0} + \prod \limits_{k=1}^\infty ~\left((1-p)\dfrac{u^k e^{-u}}{k!}\right)^{y_k}$

and this is the formula you're given.
romsek is offline  
May 12th, 2017, 10:05 AM   #3
Senior Member
Joined: Feb 2015
From: london

Posts: 121
Thanks: 0

Thanks romsek, that is a very clear explanation.
calypso is offline  

  My Math Forum > College Math Forum > Advanced Statistics

function, likelihood, liklihood

Thread Tools
Display Modes

Similar Threads
Thread Thread Starter Forum Replies Last Post
Log-likelihood function bleh Advanced Statistics 0 February 3rd, 2012 07:06 PM
Likelihood calculation AndiZ Advanced Statistics 3 February 2nd, 2011 01:47 PM
Log Likelihood~ how to get the value of ?i and ?j ? ryusukekenji Advanced Statistics 1 April 20th, 2009 03:08 PM
Likelihood and Probability a1d0ru Algebra 3 October 15th, 2008 08:39 AM
Marginalize likelihood John_Smith Advanced Statistics 0 February 6th, 2008 04:28 AM

Copyright © 2017 My Math Forum. All rights reserved.