Table 3b. Wilcoxon/Kruskal–Wallis tests on the first principal component (PC1) based on the subgroupings of the categorical clinical parameters associated with the samples.
| Metadata category | Wilcoxon/Kruskal–Wallis P-value | RANK | (n × P)/R |
|---|---|---|---|
| Disease status (case, control) | 0.00070 | 1 | 0.00490 |
| Body mass index category (normal, overweight, obese) | 0.21980 | 2 | 0.76930 |
| Family history of colorectal cancer | 0.23030 | 3 | 0.53737 |
| Diabetes history | 0.46360 | 4 | 0.81130 |
| Ever smoked | 0.74140 | 5 | 1.03796 |
| Sex | 0.86240 | 6 | 1.00613 |
| Antibiotics use | 0.93990 | 7 | 0.93990 |
Abbreviations: OTU, Operational Taxonomic Unit; PCA, principal component analysis.
PC1 was generated by PCA for the 371 OTUs that have at least one sequence assigned to them in at least 25% of the samples (Figure 5). P-values are from the Wilcoxon/Kruskal–Wallis on the first principle component. The last column shows correction for multiple hypothesis testing (Abbolito et al., 1992) using (n × P)/R, where n=total number of categorical metadata variables tested, P=raw P-value and R=sorted rank of the metadata variable.