Value of comp1 smaller than comp2 in SPLS-DA

Hi, thank you for the great package & forum! I am new to mixOmics and would like to use the SPLS-DA vignette to analyze the impact of two different antibiotics on the gut microbiome. I have followed the tutorial (tumour case study) and replicated the same analysis to my own dataset.

With a dataset of 132 samples and 162 taxonomic features, I performed performance evaluation where ncomp = 3 seems reasonable.

Next, for the optimal number of variables to include, I ran the “tune.splsda” command (with ncomp = 3, dist = “max.dist”), and 7, 90, and 100 variables are recommended for each of the components.

Now - it’s the last bit that I’m not sure:

Has anyone seen a comp1 value that’s smaller than comp2? It may be totally OK, but I’ve never seen this in PCA-type plots, where the first component always explains the most variation. If anyone has any insight or recommendation to further diagnose the data, it’d be much appreciated.

Thanks in advance!

hi @bubblepig914

Thanks for the feedback. I am referring you to that post here that should provide you with the answers!


Thank you so much for your help, Kim-Anh!