search

gabrielodom / pathwayPCA

integrative pathway analysis with modern PCA methodology and gene selection
https://gabrielodom.github.io/pathwayPCA/
11 stars 2 forks source link

Add "center" and "scale" options at object creation #37

Closed gabrielodom closed 6 years ago

gabrielodom commented 6 years ago

See Issue #36.

gabrielodom commented 6 years ago

Move discussion to Issue #38.

gabrielodom commented 6 years ago

CreateOmics() now has the argument center_scale = c(TRUE, TRUE).

gabrielodom commented 6 years ago

The center_scale parameter does not pass through to the internal svd call.

gabrielodom commented 6 years ago

For AESPCA, this was relatively easy. For SuperPCA, I still need to check that lines 149-154 of superPC_model_tStats.R does not negate the center and scale control at object creation. The loadings and PCs are calculated outside of the for() loop, so I believe that the internal center/scale to find which threshold is best will not influence how the loadings and PCs for that threshold are calculated.

gabrielodom commented 6 years ago

"The threshold is not negatively affected by the scale of the data" - our meeting.

gabrielodom commented 6 years ago

FYI: the SuperPCA_pVals() output has identical PCs, but different loading vectors, between scaled and unscaled data.