Thousands of DNA samples are typically genotyped in a GWAS, which necessitates partitioning samples into small batches of samples processed in the lab together for genotyping (e.g. the set of samples on a 96 well plate). The precise size and composition of the sample batch depend on the array and lab process used. Systematic differences among the composition of individuals in a batch (i.e. the case to control ratio or race/ethnicity of individuals on plates) and the within-plate accuracy and efficiency can result in batch effects – apparent associations confounded by batch. The problem is in essence the same problem observed with population stratification – namely that if there is an imbalance of cases and controls on a plate, and there are nonrandom (unknown) biases or inaccuracies in genotyping that differ from plate to plate – spurious associations will result.
