The receiver operating characteristic (ROC) curve was used to evaluate the predictive accuracy of genes of interest in the differentiation between samples with or without complete remission.
Finally, hierarchical clustering was performed based on Euclidean distance between AD and control samples with complete linkage method.
For samples with complete chemical analyses (excludes May 2012 samples) the average charge-balance error was 3.9%.
When considering the 135 samples with complete analyses of exons 5 8, there was no significant difference in disease-free survival between patients with or without TP53 mutations affecting the protein.
A sensitivity analysis was used to assess if there were significant difference between the sample with complete cases (n = 123) compared to the sample with imputed values (n = 129).
Two colon tissue samples with completed methylation measures were available for 262 participants (Pearson correlation between the averages of the two tissue samples = 0.66) and methylation values were available for one tissue sample for the remaining 55 study subjects.
Since estimates of PM2.5 and ozone were missing for some days, the sample size in Models C and D were smaller than those in Models A and B. Finally, Model B was replicated using the smaller dataset with non-missing air pollution measures (Model D) to check for consistency in results between the pooled sample and the sub-sample with complete air pollution information.
We select a sample with complete data as the reference.
Also, in sensitivity analyses using maximum available samples the relationships between SEP and physical performance were very similar to the relationships found in the restricted sample with complete data on all covariates (results not shown).
Simple Venn diagrams summarize the number of significant overrepresented GO categories and show the level of correlation between the Rhine and Elbe samples, with a complete list of all significantly regulated GO categories shown in Additional file 3.
