Bi-Directionally Imputing Missing Data in Gene Microarrays

• Activity Number: 320
• Type: Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Mortaza Jamshidian*+ and Amol Kumar
• Companies: California State University, Fullerton and California State University, Fullerton
• Keywords: Gene Microarray ; Missing data ; Imputation ; genes ; tissues

Abstract

We obtained the gene microarray data from Alizadeh and Yoshimoto to compare modified imputation techniques. 10%, 20%, 25% and 30% missing data was introduced randomly into the complete portions of the data sets and after imputing we computed a normalized Frobenius norm and the correlation between the imputed data set and the complete data set. K-nearest neighbors, principal components analysis and normal distribution based imputation were considered. We sought improvements by modifying current techniques; in particular we found that imputing a microarray its transpose and taking the average of the results may yield improvements; we call this method bi-directional imputation. For methods which require a covariance matrix when there are more variables than observations we used a shrinkage estimator.