Vol: 42 Page: 103

Authors:
Alicja Szabelska
Joanna Zyprych-Walczak
Idzi Siatkowski

Title:
 Application of linear mixed models in the selection of genes from microarray experiments with repeated measurments

Language: English

Keywords:
linear fixed model
linear mixed model
R software
selection
permutational F test
classification

Summary:
This paper focuses on the application of linear mixed models to microarray experiments. The main focus is on experimental design with biological as well as technical replicates. The results suggest that, depending on the considered number of top genes, different tests for linear fixed model or linear mixed model show better outcomes. In particular, cross validation revealed that the fixed model with parametric tests along with the mixed model with permutational tests based on residuals attained the lowest classification errors. On the other hand, ROC curve analysis implied that parametric tests for fixed as well as mixed model return the highest values for performance effectiveness.

Display