S231 Lecture 4 Cyntha

# S231 Lecture 4 Cyntha

This preview shows page 1. Sign up to view the full content.

This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: infected. Statistics and Actuarial Science () Model Fitting, Estimation and Checking Jan 2013 19 / 45 HIV example. Suppose Y represents the number of persons infected with the human immunode...ciency virus (HIV) in a randomly selected group of n persons. Modelled by Binomial P (Y = y ; ) = f (y ; ) = n y (1 y )n y for y = 0, 1, . . . , n where is the fraction of the population that are infected. In a random sample of n persons tested for HIV, y = y is the observed number infected. n L( ) = (y ) y (1 ^ = y /n. )n y for 0 < < 1 is maximized by the value Statistics and Actuarial Science () Model Fitting, Estimation and Checking Jan 2013 19 / 45 HIV example. Suppose Y represents the number of persons infected with the human immunode...ciency virus (HIV) in a randomly selected group of n persons. Modelled by Binomial P (Y = y ; ) = f (y ; ) = n y (1 y )n y for y = 0, 1, . . . , n where is the fraction of the population that are infected. In a random sample of n persons tested for HIV, y = y is the observed number infected. n L( ) = (y ) y (1 ^ = y /n. )n y for 0 < < 1 is maximized by the value If n is large and is very small, we might prefer to model Y using the Poisson() distribution with = n. Why is this reasonable? Statistics and Actuarial Science () Model Fitting, Estimation and Checking Jan 2013 19 / 45 If you' a likelihood value, (relative) size matters: re The shape of L( ) and its maximum value are not aected if we multiply L( ) by a constant. Relative value at two dierent values of the parameter, e.g. L( 1 )/L( 2 ) is important. It determines how much more or less consistent the data is with the parameter 1 versus 2 . Statistics and Actuarial Science () Model Fitting, Estimation and Checking Jan 2013 20 / 45 If you' a likelihood value, (relative) size matters: re The shape of L( ) and its maximum value are not aected if we multiply L( ) by a constant. Relative value at two dierent values of the parameter, e.g. L( 1 )/L( 2 ) is important. It determines how much more or...
View Full Document

## This note was uploaded on 04/17/2013 for the course STAT 231 taught by Professor Cantremember during the Winter '08 term at Waterloo.

Ask a homework question - tutors are online