The obtained density distribution for the p-values of all genes in one selected tissue, ρ(p), meets the normalization condition ∫ 0 1 ρ p · d p = 1.
We considered that the conditions are met when the normalization was done on sample of a size higher or equal than 30 individuals for each SNP (the principle of the central limit theorem).
where η is normalization factor to meet the power constraint at the eNB.
All standardized expression data were then log-transformed to normalize the residuals in order to meet the necessary assumptions for the statistical tests (after normalization, the residuals did meet these assumptions).
Since two normalization methods were applied, qualified probes/genes from either method that meet the above criteria were included in the candidate gene lists.
Only features that met the following criteria were called normalization features (NF) and were used to calculate the scale factor: only Bb features with less than 10% saturated pixels; median signal more than six fold above background and observed in all 6 arrays.
Across all patients, 38.1% did not meet normalization criteria during the study, 19.0% transiently normalized (met criteria at end of study phase 1), and 23.8% did not score in the normal range after study phase 1 but did at study endpoint.
A total of approximately 1000 performance measures were extracted by selecting those that met the different combination of parameter values after the normalization process for each experiment.
If subjects met the inclusion criteria, they were studied on two occasions: 1) an intravenous glucose tolerance test (IVGTT) after acute glucose normalization and 2) an IVGTT without acute glucose normalization.
