|
Activity Number:
|
69
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #307106 |
|
Title:
|
Estimation for Finite Mixture Multinomial Models
|
|
Author(s):
|
Nagaraj Neerchal*+ and Minglei Liu and Jorge Morel
|
|
Companies:
|
University of Maryland Baltimore County and Medtronic, Inc. and Procter & Gamble
|
|
Address:
|
Department of Math and Statistics, Baltimore, MD, 21250,
|
|
Keywords:
|
mixture model ; Fisher's scoring algorithm ; EM algorithm ; multinomial
|
|
Abstract:
|
Mixture of multinomial distributions is important for both theoretical and practical reasons. Because of the specialty of this model, it is not easy to get the maximum likelihood estimates of the parameters. Several methods, including Fisher's Scoring algorithm and EM algorithm are available in the literature to get the MLE numerically. The authors consider an approximate of the information matrix of the mixture of multinomial model and proposes an Approximate Fisher's Scoring algorithm. We also investigate the properties of this algorithm and compare it with the other well known algorithms by simulation.
|
