IMA Complex Systems Seminar

2:30 Wednesday, October 29, 2003

 

 

Missing Value Estimation Methods for DNA Microarrays

 

Shmuel Friedland

Department of Mathematics, Statistics, and Computer Science

University of Illinois at Chicago

Chicago, IL  60607-7045

 

In this talk we first survey the known methods for estimation for missing values in gene microaarays:

 

1.  Row average.

2.  Clustering methods.

3.  Singular value decomposition (principal-component analysis).

 

The draw back of these methods is that the recovery of the missing data is done independently, i.e. the completion of each missing value does not influence the completion of other values.

 

Next we discuss a new method for a completion of missing values in which the completion is done simultaneously.  It is inspired by the methods for solutions of the inverse eigenvalue problems.