Table 1.
Overview of the workflow and main functions in the ‘triplot’ R package.
| Function | Description |
|---|---|
| First layer: Latent variable (LV) modeling | |
| Custom a | Perform LV modeling of high-dimensional (metabolomics) data. |
| makeTPO() a | Initiate a triplot object (TPO) from LV model |
| Second layer: Correlations | |
|
makeCorr()
or custom b |
Perform correlation analysis between LV observation scores and exposures or covariates. |
| addCorr() | Add correlation results to the TPO. |
| Third layer: Associated risk | |
|
crudeCLR(),
crudeLR(), or custom c |
Calculate risk associations (i.e., odds ratio or hazard ratio) in (conditional) logistic regression or association with intermediate risk markers (i.e., beta coefficient) in linear regression. |
| addRisk() | Add risk associations to the TPO. |
| Visualizations | |
| checkTPO() | Generate a heatmap visualizing correlations and risk associations to identify relevant LVs for the triplot visualization. |
| triplot() | Create a triplot containing LV analysis results, correlations, and risk associations. |
a Actual LV modeling is purposely omitted from the triplot package to give the user the choice of LV method, such as PCA, FA, or PLS. The makeTPO() function will accept any input that conforms to scores and loadings. b makeCorr() constitutes a convenience function for standard correlation analysis (Pearson, Spearman, Kendall). Partial correlation requires custom scripts and is covered in the tutorial. c crudeLR() and crudeCLR() constitute convenience functions for (conditional) logistic regression. Adjusting associations for covariates requires custom scripts and is covered in the tutorial.