Summary¶

Basically, we create an elliptical envelop with respect to the covariance matrix and a specified level of contamination (0.05). This was done with the first two components for two different lower dimenensional projections of the expression data. The envelop implementation used was [Rousseeuw99].

The results of the outlier analysis. The PCA results were used because the results seemed both more consistent and made more appropriate outlier selections with respect to the density plots.

• sample-outliers.csv

To determine the sample name for a given studie’s outlier look at the covariates file and use the index reported in the above file. The covariate files are located after each plot.

Decision boundries¶

• GSE19187-1-cov.csv

• GSE18965-1-cov.csv

• GSE8190-cov.csv

• GSE43696-1-cov.csv

• unpublished-cov.csv

• GSE51392-1-cov.csv

• GSE67472-1-cov.csv