Hi all,
I’m analysing microbial community data from a factorial experiment with a randomised block design. Treatments were assigned within spatial blocks (Block is a random factor), and samples were destructively collected across multiple time points. Each plot (Plot_ID) was sampled independently at each time, so there are no true repeated measures.
Each Plot_ID × Time combination includes multiple technical replicates (e.g., three samples per plot). These were not averaged but kept as separate entries in the dataset.
I’m using the adonis2 function with a Bray-Curtis distance matrix. I’ve seen the following model used:
adonis2(dist ~ Treatment * Time, data = meta, strata = meta$Block)
My questions are:
1. Is it scientifically valid to use strata = Block without including Block in the formula, given that Block is a random factor?
2. Would including + Block in the formula be statistically inappropriate in this context?
3. Should technical replicates per Plot_ID be accounted for via strata = Plot_ID instead?
4. If both Block and Plot_ID are relevant, how should permutation structure be handled in adonis2 (e.g., nested strata or alternative approach)?