Assessment of Weighted KNN Imputation and Multiple Imputation Techniques Using Colorectal Cancer MiRNA Data

• Activity Number: 457
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Anvar Suyundikov*+ and John Stevens and Marty Slattery and Roger Wolff and Jennifer Herrick
• Companies: and Utah State University and University of Utah and University of Utah and University of Utah
• Keywords: missingness ; weighted KNN imputation ; multiple imputation ; differential expression ; miRNA

Abstract

High-dimensional bioinformatics data often suffer from missing values due to various experimental and technical reasons. The statistical analyses of missing data may lose power and have biased inference. In this presentation, we demonstrate the strengths and weaknesses of the weighted KNN imputation and multiple imputation techniques over the case deletion technique using a large colorectal cancer (CRC) dataset. This CRC dataset contains extensive lifestyle, genetic, survival, and tumor marker data collected from the study participants. Differential expression tests of miRNAs are performed using various statistical methods while considering the correlation structure in the imputed data and controlling for additional risk factors by including them as covariates.