Activity Number:
|
15
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, August 11, 2002 : 2:00 PM to 3:50 PM
|
Sponsor:
|
ENAR
|
Abstract - #301868 |
Title:
|
Bio-informatics Analysis in Dichotomous Genetic Marker Information
|
Author(s):
|
Kuang-Ho Chen*+
|
Affiliation(s):
|
National Taiwan University
|
Address:
|
Room 1540, Basic Med. Bld. 1, Sec 1 Jen-Ai Road, Taipei, International, 100, Taiwan
|
Keywords:
|
exact logistic regression ; existence with binomial maximum likelihood ; screenout non-informative genetic markers ; mathematical screen-out ; statistical inference for screen-in ; huge data small samples
|
Abstract:
|
Any available number T of tumor tissues is very small in comparing with the number G of genetic markers, T < < G. A mechanism is in demand that would screen out non-informative information generated from the genetic markers. Among others, the LR (logistic regression) with maximum binomial likelihood (MBL) is recommended, because the proposed necessary and sufficient existence condition for K binary explanatory variables is that none is 0 for the 2C summarized statistical counts, where C is 2 to the power of K. For K pieces of dichotomous information generated from G genetic markers, C could be as large as 2 to the power of L, where L is 2 to the power of K after counting on all levels of interactions. The proof of the critical condition is simple and affords the constructing procedure of the exact coefficients to be in closed form for K=1 and direct numerical iteration for K > 1. The proposed procedure of exact logistic regression with MBL (XLRMBL) jointed with the procedure for conditional probability of Fisher exact test (CPFXT) ascertains the function of a genetic marker in clinics or pathological diagnosis.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2002 program |