I have been diving deep into compositional data analysis. Here is a great thread containing a conversation with Thomas Quinn, who is pushing this type of research in microbiology, about some intricacies of compositional data analysis . In that thread proportionality is defined but essentially it can be used as a stand-in replacement for correlation that is compositionally valid.

My question is out of scope for that thread so I thought of asking here. I have a compositional dataset X that is made up of genes as components and samples as compositions. I've made a transformation (that might not be the correct term) where I sum up the counts for different groups of genes. However, a single component sometimes belong to multiple groups so it's not compositional anymore. Furthermore, I normalize by the length of the genes in each group so the data is certainly not compositional.

My question is whether or not I can use the proportionality rho metric in this case?

Certainly I can use it and get an answer out with propr but would this be valid after I've applied these types of data manipulation?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.