Yes - I think the one exception is the first time you study a phenotype as black swans may be found. When that not found (unfortunately they are rare). So suggests standard nested - from initial small core multi-modal, then larger and lighter clinical, then online remote sampling population level.
Gerome Breen
<big sigh>. As a rule of thumb, if you gather more multi-modal data, you are going to need more samples (patients) rather than less to be confident of results. Importantly you need some sort of credible belief that for the measurements you propose to do you have some hope of the analysis working
