Multivariate & Dimensionality
Multivariate figures are the most parameter-sensitive figure type in the Publishable Figs library. A heatmap with a different clustering method, a PCA with different scaling, or a dendrogram with a different distance metric can produce visually completely different results from the same data. Every parameter must be disclosed — and every parameter choice must be justified.
Parameters are not cosmetic. The choice of distance metric, linkage method, and color scale in a heatmap changes which clusters appear, which genes appear co-regulated, and which groups appear similar. A reviewer who cannot reproduce your heatmap from the disclosed parameters cannot evaluate your interpretation.
How to Choose
- Heatmap: show expression or intensity patterns across features and samples
- Parallel coordinates: compare multiple continuous variables across groups or cases
- PCA biplot: show both sample scores and variable loadings in reduced dimensions
- Hierarchical clustering: show group structure without a priori group assignment
Figure Types
Heatmap
Clustering method, distance metric, scaling, and color scale all disclosed — mandatory for any heatmap.
Parallel Coordinates
Multi-variable profile comparison with axis normalization, group coloring, and missing data handling.
PCA Biplot
PC scores + loadings with variance explained per PC and centering/scaling disclosure.
Hierarchical Clustering
Dendrogram with linkage method, distance metric, and cut-height or optimal k selection.