Dear MixOmics developers,
My name is Johannes and I’m a postdoc working on metagenomics data.
Many thanks for developing DIABLO and mixOMICS. I’m really looking forward trying it out!
What would you say is the lower end in terms of the number samples for DIABLO to do ok? I see that you have analyzed data with ~200 samples which is far less than our model. To asses model performance, do you need to split up your data for cross validation? Do you think DIABLO would be able to do robust inference on less than 100 samples and if so, would I need to increase regularization (if that’s possible)?