Genes with similar expression patterns were grouped as hierarchical clusters with distances between samples computed using Pearson correlations for similarity measures and average linkage as the clustering algorithm.
In a causal filter (§5.3), each output sample is computed using only current and past input samples--no future samples.
Posterior conditional probabilities of the component membership for each sample were computed using the fitted parameters, and a threshold of ≥0.5 was used to classify a sample as having positive expression of the receptor.
A total of 13 attributes per image sample were computed using color and texture models: HSI (hue, saturation, and intensity) color histogram statistics and GLCM probabilities.
The OTU richness rarefaction curve of each sample was computed using a self-written function in R (Supplementary Data S2).
Enrichment of GO terms in 127 genes that had STAT1 and NF-κB-p65 binding sites and greater than 2-fold enrichment of Pol II binding in the co-treated sample was computed using DAVID (Jiao et al., 2012).
In addition, the required sample size computed using the effect size based on the score allowed obtaining the expected power, whatever the method of analyse of the data (score or Rasch model).
Moreover, the required sample size computed using the effect size on the latent trait did not allow obtaining the expected power (78.9% instead of 90%), stressing the fact that the usual formula might not be well-suited for IRT models.
Next, the eigenvectors u iw of the scatter matrices of each class sample were computed using the efficient technique proposed in [ 15], by first solving the eigenvalue problem: (3) where λ jω was the j th eigenvalue, and v jω is the j th eigenvector of matrix.
Fragment analysis was performed with Gene Mapper software (Life Technologies, USA) and data was analyzed using Coffalyser software (MRC, Holland) where DNA copy number ratios of tumor samples were computed using the matched blood sample.
